Optimal Simultaneous Production of Hydrogen and Liquid Fuels from

Article pubs.acs.org/IECR

Optimal Simultaneous Production of Hydrogen and Liquid Fuels from Glycerol: Integrating the Use of Biodiesel Byproducts Mariano Martín*,† and Ignacio E. Grossmann‡ †

Department of Chemical Engineering, Universidad de Salamanca, Pza. Caídos 1-5, 37008 Salamanca, Spain Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213, United States

‡

ABSTRACT: In this paper, we present the optimization of the production of hydrogen and/or liquid fuels from glycerol. We propose a limited superstructure embedding a number of alternative technologies. Glycerol is first reformed using either aqueous phase reforming, steam reforming, or autoreforming. The gas obtained is cleaned, and its H2 to CO ratio is adjusted (bypass, PSA, and/or water gas shift). Next, the removal of CO2 is performed by means of PSA, and the syngas is fed to the Fischer− Tropsch reactor. The products obtained are separated while the heavy fraction is hydrocracked. The optimization of the system is formulated as a mixed integer nonlinear programming (MINLP) that is solved first for the optimal production of hydrogen alone and next for the simultaneous production of liquid fuels and hydrogen. The production of hydrogen is competitive with that obtained from switchgrass as long as the glycerol price is below $0.05/lb ($0.110/kg) using aqueous phase reforming. For the liquid fuels to be attractive, the glycerol price must be below $0.025/lb ($0.055/kg) using autoreforming. The option of integrating this facility with a biodiesel one is promising from an economical point of view. biodiesel production requires energy,3 and the production of fuels out of a byproduct can help operate the facility reducing its dependency on nonrenewable sources. Hydrogen was first proposed as an energy carrier by Cecil.8 Current developments on fuel cell technology for both stationary generation of electricity and road transportation represent an important step toward energy security, where the hydrogen economy becomes key for the feasibility of the technology.6 However, as any other alternative fuel, the availability and low cost of fossil fuels has slowed their development.9,10 On the other hand, the Fischer−Tropsch synthesis has been used in moments of difficult access to crude oil to produce synthetic gasoline and diesel from coal, and it is now spreading to biomass and other sources of syngas.11,12 In order to generate syngas from glycerol, a number of recent studies evaluate the reforming of glycerol.13−16 However, the gas resulting from this stage has to be further purified.17−22 In this paper, we focus on the simultaneous production of FT-diesel and hydrogen from glycerol so as to increase the yield from vegetable, algae, or cooking oil to diesel substitutes. Because the raw material, glycerol, comes from the biodiesel industry, we target the production of FT-diesel so that both are similar biofuels. To improve the design and the energy efficiency as well as to decide whether it is interesting to produce hydrogen, FT-liquid fuels, or both, mathematical optimization techniques are used.23 We propose a limited superstructure optimization approach where we first construct a flowchart embedding the various process units involved in hydrogen and synthetic liquid fuels production, considering various technological alternatives for some of the processes.

1. INTRODUCTION The use of biomass to obtain liquid fuels has attracted interest due to their compatibility with the current gasoline and diesel supply chains. However, the profitability of biofuels depends heavily on the economics of the byproducts. This is the case of corn-based ethanol, where the dried distilled grains with solubles (DDGS) provide an important credit1 or the lignocellulosic-based ethanol in which the hydrogen produced is a major asset.2 In the case of the biodiesel obtained from the transesterification of oil (vegetable oil from seeds, cooking, or algae oil3), glycerol is a valuable byproduct that can contribute to the profitability of the biodiesel production process due to the large number of compounds that can be produced or manufactured from it and their many potential applications.4,5 Figure 1 shows the distribution of the use of glycerol as an

Figure 1. Distribution of products currently obtained from glycerol.

ingredient for different products.4 The expected increase in the amount of glycerol produced resulting from the production of biodiesel presents a new scenario with the need to find new markets. Among them, we can consider the production of fuels. Among these fuels, hydrogen6 and FT-fuels7 are promising due to their straightforward use within the biorefinery. Furthermore, © 2014 American Chemical Society

Received: Revised: Accepted: Published: 7730

January 6, 2014 March 14, 2014 April 10, 2014 April 10, 2014 dx.doi.org/10.1021/ie500067d | Ind. Eng. Chem. Res. 2014, 53, 7730−7745

Industrial & Engineering Chemistry Research

Article

Figure 2. Schematic flowchart for the production of hydrogen and FT-fuels.

Figure 3. Details of the flowchart for the alternatives for glycerol reforming.

2. OVERALL PROCESS DESCRIPTION The process consists of five different parts. The first one is the reforming of the glycerol. Three different technologies are evaluated: (1) steam reforming (SR), (2) autoreforming (AR), and (3) aqueous phase reforming (APR). The two first generate raw syngas operating at high temperature, while the last one generates only H2 and CO2 at a much lower temperature, but it requires a large amount of liquid water. Next, the last traces of hydrocarbons are removed in a PSA system with a bed of silica gel. After reforming, the molar ratio of H2 to CO may need to be adjusted to a H2/CO from 1 to 2 according to the results by Wang et al.24 or else hydrogen is obtained and purified. In order to do so, a water gas shift reactor, bypass, and hybrid membrane/pressure swing adsorption (PSA) with a bed of oxides for H2 purification are considered. The split fraction depends on the performance of the reforming stage.

The particular feature is the modeling effort to obtain equationoriented models for the most important equipment, in particular glycerol reformers, from experimental data so as to predict the optimal operating variables, such as temperature, oxygen, and/or steam added or composition of the raw syngas. These units are interconnected to each other through network flows and other utility streams. The goal is to optimize the structure and the operating conditions to maximize the profit in the production of synthetic fuels. The optimization of the system is formulated as a mixed-integer nonlinear programming (MINLP) problem, where the model involves a set of constraints representing mass and energy balances for all the units in the system. We then perform heat integration of the resulting process followed by an economic evaluation to compare the different options. Finally, the integration of this facility into an algae-based biodiesel plant is considered to determine its effect on the economics of such complex. 7731

dx.doi.org/10.1021/ie500067d | Ind. Eng. Chem. Res. 2014, 53, 7730−7745


Article

The third step corresponds to the removal of sour gases, CO2. According to the study by Martı ́n and Grossmann2, a PSA system with a bed of Zeolite 5A is recommended. Once the gas is purified, the Fischer−Tropsch synthesis is carried out. Over a catalyst of iron or cobalt, the synthesis gas consisting of H2 and CO undergoes a series of chemical reactions where it is transformed into liquid hydrocarbons. The optimal conditions (ratio of H2 to CO and the working temperature at the reactor) are to be optimized. The working pressure is assumed to be 30 bar according to experimental results.24 Synthetic diesel is produced together with gasoline, heavy fuels (FT-liquids), and gas products.25 Finally, the FT-liquids produced are separated from the gas products and fractionated using an atmospheric distillation column. The heavy oil is then hydrotreated in order to increase the yield toward diesel and gasoline. The results from Bezergianni et al.26 are used to develop an input−output model for the hydrocraking of the heavy oil. Figure 2 shows the flowchart embedding the different alternatives.

Q(Furnace) = fc(Glycerol,Src1,Furnace) × (c p(Glycerol) × (T(Furnace,Ref) − T_amb) + fc(H 2O,Src1,Furnace) × (c p(H 2O) × (T(Furnace,Ref) − T_amb)

(2)

3.1.1. Autoreforming. The chemical reactions taking place are of the form given by eq 3. The experimental data shown in Douette’s paper14 are used to obtain a surrogate model using a design of experiments approach for predicting the gas composition exiting the reformer as a function of the temperature and feed composition (steam and oxygen added). Due to the nature of the model, the bounds for the operating variables are set to those given by the experimental data available so as to obtain a reliable model (eqs 4−9). C3H8O3 + x H 2O + yO2 → aCO + bH 2 + cCH4 + dCO2 + e H 2O + ...

3. MATHEMATICAL MODELING The operation of all the units in the production process of liquid fuels and hydrogen from glycerol is modeled using surrogate models, design equations, and mass and energy balances. The superstructure is written in terms of the total mass flows, component mass flows, component mass fractions, pressures, and temperatures of the streams in the network. The components in the system belong to the set J = { H2O, glycerol, gasoline, diesel, heavy, CO2, CO, O2, N2, H2, CH4, C2H2, C2H4, methane, butane, propanediol, C}. The different units in the superstructure are modeled as described below. For the sake of reducing the length of the paper, we refer to previous papers by the authors for the detailed models of particular equipment related to gas treatment and FT-synthesis.2,12,27 3.1. Reforming. In this paper, we consider three of the most promising reforming technologies to obtain a gas phase from glycerol. The first one is steam reforming, an endothermic process with high yield to hydrogen. The second one is autoreforming, a process that combines steam reforming and partial oxidation so that the oxidation of part of the glycerol provides the energy required for the steam reforming reactions.10 Finally, we consider aqueous phase reforming, a technology that produces only H2 and CO2 using water in the liquid state.28 In order to develop equation-oriented models, experimental data from the literature are used. The glycerol is fed to a furnace to heat it and, if required, to gasify it (Figure 3). Equation 1 is used for steam and autoreforming, and eq 2 is used for aqueous phase reforming.

(3)

0 ≤ Oxygen_add ≤ 0.8

(4)

1 ≤ Steam_add ≤ 3

(5)

700°C ≤ T(Furnance,Ref) ≤ 1000°C

(6)

fc(O2 , Src1/2,Furnace) = Oxygen_add ×

fc(Glycerol,Src1,Furnace) × 3 × MW(O2 ) MW(Glycerol) (7)

fc(H 2O,Src2,Furnace) = Steam_add

fc(Glycerol,Src1,Furnace) × 3 × MW(H 2O) MW(Glycerol) (8)

fc(J,Src2,Furnace) + fc(J,Src1,Furnace) + fc(J,Src3,Furnace) = fc(J,Furnace,Ref)

(9)

The product gas composition is calculated using the correlations developed from Douette’s data,14 given by eqs 10−13 nH2 =

Q(Furnace) = fc(Glycerol,Src1,Furnace) × (c p(Glycerol)

fc(Glycerol,Src1,Furnace) × (0.0909 MW(Glycerol)

× (Tb(Glycerol) − T_amb)

+ 0.1018 × Oxygen_add + 0.1847 × Steam_add

+ dH_vap_0(Glycerol)

+ 7.9854 × 10−3 × T_reforming

+

TFurnace

∫Tb(Glycerol) cp_v(j)dT )

+ fc(H 2O,Src2,Furnace) + fc(O2 ,Src2,Furnace)

+ 0.2262 × Oxygen_add × Steam_add TFurnace

∫T_amb TFurnace

∫T_amb

− 9.0133 × 10−3 × Oxygen_add × T_reforming c p_v(j)dT )

− 5.2520 × 10−4Steam_add × T_reforming − 6.7076 × 10−5Oxygen_add × Steam_add

c p_v(j)dT )

× T_reforming)

(1) 7732

(10)



Article

Figure 4. Model fit for autoreforming. Moles of component per mol of glycerol.

data within the experimental range of data especially for the main species of the raw syngas. In the case of the CH4, we must point out the fact that we are dealing with traces of the species, and thus, even though the agreement is not that good, the error is within 20% in most of the cases. The atomic balances must also hold

fc(Glycerol,Src1,Furnace) (− 0.4502 − 1.7314 × Oxygen_add MW(Glycerol)

nCO =

− 0.4162 × Steam_add + 3.9189 × 10−3 × T_reforming + 1.1866 × Oxygen_add × Steam_add − 1.1920 × 10−3 × Oxygen_add × T_reforming − 1.0400 × 10−4 Steam_add × T_reforming − 7.6660 × 10−4Oxygen_add × Steam_add

nCO2 =

H: (2 × Steam_add × 3 + 8)

(11)

× T_reforming)

fc(Glycerol,Src1,Furnace) MW(Glycerol)

− (2 × nH2 + 4 × nCH4 + 2 × nH2O + 2 × nC2H2

fc(Glycerol,Src1,Furnace) (0.3906 + 0.2149 × Oxygen_add MW(Glycerol)

+ 4 × nC2H4) = 0

+ 0.8711 × Steam_add + 8.6620 × 10−4 × T_reforming + 0.4791 × Oxygen_add × Steam_add + 1.3223 × 10−3

C: (3)

× Oxygen_add × T_reforming − 8.1260 × 10−4 × Steam_add × T_reforming − 7.2680 × 10−4Oxygen_add × Steam_add

(14)

fc(Glycerol,Src1,Furnace) − (nCO + nCO2 + nCH4 + nC MW(Glycerol)

+ 2 × nC2H2 + 2 × nC2H4) = 0

(15)

(12)

× T_reforming)

O: (2 × Oxygen_add × 3 + 3 + Steam_add × 3) nCH4 =

fc(Glycerol,Src1,Furnace) (−3.726 × 10−8 MW(Glycerol)

fc(Glycerol,Src1,Furnace) − (nCO + 2 × nCO2 + 2 × nO2 MW(Glycerol)

− 3.773 × 10−8 × Oxygen_add − 3.108 × 10−7 × Steam_add

+ nH2O) = 0

+ 4.317 × 10−5 × T_reforming − 1.445 × 10−7

The outlet stream is given by

× Oxygen_add × Steam_add + 1.002 × 10−5 × Oxygen_add

fc(J,Ref,HX1) = nj × MW(J);

× T_reforming + 9.767 × 10−6 × Steam_add × T_reforming −5

− 2.453 × 10

(16)

× Oxygen_add × Steam_add × T_reforming)

∀j

(17)

Finally, we assume that the reformer operates adiabatically

(13)

Q react =

The fitting of the model is shown in Figure 4. As shown, good agreement is found between eqs 10−13 and the experimental

∑ ΔHjf |T(furnace) j

7733

(18) dx.doi.org/10.1021/ie500067d | Ind. Eng. Chem. Res. 2014, 53, 7730−7745

Industrial & Engineering Chemistry Research Q prod =

Article

∑ ΔHjf |T(furnace)

reformer. Due to the complexity of the profiles for the gas generated at the reformer, eqs 24−27 are also complex. However, the fitting of the experimental data with these equations is satisfactory as shown in Figure 5. We focus our analysis within the range of the experimental values used for the variables, temperature (327−727 °C) and ratio of steam added (up to 9 mol of water per mol of glycerol), so that the model is reliable

(19)

j

Q prod ≈ Q react

(20)

3.1.2. Steam Reforming. The reactions taking place are of the form given by eq 21. We model the operation of this unit using the results presented by Adhikari et al.13 The figures shown in that paper are used to obtain the data for the gas composition as a function of the temperature and the steam added. The profiles of the gas composition as a function of the temperature and the added steam are complex, and hence, simple correlations are not enough to predict the outlet gas. A two-step procedure is used to correlate the outlet gas as a function of the temperature and the added steam. First, for each of the values of added steam, we correlate the outlet gas (CO, CO2, H2, and CH4) as a function of the temperature. Next, we include the effect of the ratio of steam added so as to develop an equation-based model to predict the gas composition as a function of the temperature and the steam injected in the nH2 =

C3H8O3 + nH 2O → aCO + bH 2 + cCO2 + dCH4 + ... (21)

fc(H 2O,Src2,Furnace) = Steam_add

fc(Glycerol,Src1,Furnace) MW(H 2O) MW(Glycerol) (22)

fc(J,Src2,Furnace) + fc(J,Src1,Furnace) = fc(J,Furnace,Ref)

(23)


⎛−1.12 × 10−11(Steam_add3) + 1.948 × 10−10(Steam_add2)⎞ ⎜ ⎟ × (T_reforming + 273)4 )+ ⎜ ⎟ −10 −10 ⎝−8.677 × 10 Steam_add + 7.681 × 10 ⎠ ⎛ 3.5751 × 10−8(Steam_add3) − 6.1754· 10−7Steam_add2 ⎞ ⎜ ⎟ × (T_reforming + 273)3 + ⎜ ⎟ −6 −6 ⎝+2.7097 × 10 Steam_add − 2.4456 × 10 ⎠ ⎛−4.2181 × 10−5(Steam_add3) + 7.2376 × 10−4(Steam_add2)⎞ ⎜ ⎟ × (T_reforming + 273)2 + ⎜ ⎟ ⎝−3.1299 × 10−3Steam_add + 2.8955 × 10−3 ⎠ (0.021807 × (Steam_add3) − 0.37172 × (Steam_add2) + 1.5866 × Steam_add − 1.5026) × (T_reforming + 273)+ ( −4.171 × Steam_add3 + 70.653 × Steam_add2 − 298.11 × Steam_add + 288.16))

nCO =

(24)

fc(Glycerol,Src1,Furnace) (7.3154 × 10−11 × Steam_add MW(Glycerol) − 7.1134 × 10−10) × (T_reforming + 273)4 + (− 2.3117 × 10−7Steam_add + 2.2197 × 10−6) × (T_reforming + 273)3 + (2.6955 × 10−4 Steam_add − 2.5493 × 10−3) × (T_reforming + 273)2 + (− 0.1377 × Steam_add + 1.2817) × (T_reforming + 273) + (2.6028 × 10 × Steam_add − 2.3855 × 102)

nCO2

(25)

3 2⎞ −8 −5 ⎛ fc(Glycerol,Src1,Furnace) ⎜−2.1725 × 10 (T_reforming + 273) + 4.3132 × 10 (T_reforming + 273) ⎟ = ⎜ ⎟ MW(Glycerol) ⎝−0.0272 × (T_reforming + 273) + 6.3326 ⎠

⎛ 6.628 × 10−11(T_reforming + 273)4 − 1.9627 × 10−7(T_reforming + 273)3 ⎞⎞ ⎟⎟ × (Steam_add)⎜⎜ ⎟⎟ 2 −4 ⎝+2.1633 × 10 (T_reforming + 273) − 0.10479 × (T_reforming + 273) + 19.06 ⎠⎠

7734

(26)



Article

Figure 5. Model fit for the steam reformer. Moles of component per mol of glycerol.

nCH4 =


⎞ ⎛ ⎛− 9.7452 × 10−6Steam_add3 ⎞ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎜+ 2.0612 × 10−4Steam_add2 ⎟ ⎟ ⎜ ⎟ × (T_reforming + 273) (− 0.4790 × log(Steam_add) + 8.1713) + ⎜ ⎟ ⎜ ⎜− 1.3085 × 10−3Steam_add ⎟ ⎟ ⎜ ⎟⎟ ⎜⎜ ⎟ ⎜ ⎠ ⎝− 3.3209 × 10−3 +⎟ ⎜ −5 −4 −3 2 1 exp( (T reforming 273) (5.1956 10 Steamadd 2.7429 10 Steam add 2.8799 10 )) + − _ + × × − × _ + × ⎟ ⎜ ⎟ ⎜ 3 2 ⎟ ⎜ (0.0155 × Steam_add − 0.3021 × Steam_add + 1.7829 × Steam_add − 4.8415) ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎠ ⎝

The atomic balances must also hold (eqs 28−30) H: (2 × Steam_add + 8)

O: (3 + Steam_add)


fc(Glycerol,Src1,Furnace) − (nCO + nCO2 + nCH4 + nC MW(Glycerol)

+ 2nC2H2 + 2nC2H4) = 0

(30)

To avoid obtaining negative values for CH4, we define the stream flows as positive variables, avoiding operating conditions that may result in negative flows for CH4. In the case of H2, this fact is not likely to happen because we aim for maximum production of hydrogen. The outlet stream is given by

(28)

C: (3)


− (nCO + nCO2 + 2nO2 + nH2O) = 0

− (2 × nH2 + 4nCH4 + 2nH2O + 2nC2H2 + 4nC2H4) = 0

(27)

fc(J,Ref,HX1) = nj × MW(J );

(29) 7735

∀j

(31)



Article

yCH = 0.04 + 4.5 × 10−4 × (T(Furnace,Ref) − 225)

T_reforming = 0.5 × (T(Furnace, Ref) + T(Ref,HX1))

4

(32)

(38)

We assume adiabatic operation of the reformer Q react =

yC H = 5 × 10−5(T(Furnace, Ref) − 225)

∑ ΔHjf |T(furnace)

2

(33)

j

Q prod =

(34)

Q prod = Q react

(40)

(35)

3.1.3. Aqueous Phase Reforming. Recently, another technology has been proposed that allows the production of hydrogen from glycerol directly. It is called aqueous phase reforming. In this process, the water is kept in liquid state by increasing the operating pressure. Using the experimental results from Shabaker et al.,28 we propose a simple linear model for this reformer based on only two pair of data that are available, where the gas molar fraction depends on the temperature within the range of experimental values. yH = 0.66 − 5 × 10−4(T(Furnace,Ref) − 225)

(36)

yCO = 0.3

(37)

2

2

fracj =

(fracCgas ×

fc(Glycerol,Furnace,Ref) × MW(C) × 3 MW(Glycerol)

xj =

yj × MW(j) ∑j yj × MW(j)

(42)

⎛ MW(C) ⎞ ⎛ MW(C) ⎞ mC = ⎜ ⎟xCH ⎟xCO2 + ⎜ ⎝ MW(CO2 ) ⎠ ⎝ MW(CH4) ⎠ 4 ⎛ MW(C) ⎞ + ⎜2 × ⎟x C H MW(C2H6) ⎠ 2 6 ⎝

⎛

) × ⎝x × ( ⎜

MW(j) ; nCj × MW(C)

fc(H 2,Ref,HX1) = fc(CO2 ,Ref,HX1) ×

j

nCj × MW(C) MW(j)

)⎞⎠ ;

x H2 (46)

In order to meet the mass balance, we assume that in the liquid phase ethanol, propanediol, and methanol (in this order of amount) are the most important byproducts.28 ⎛ 2 × fc(EtOH,Ref,HX1) ⎞ ⎛ fc(MetOH,Ref,HX1) ⎞ ⎜ ⎟+⎜ ⎟ MW(EtOH) ⎝ ⎠ ⎝ MW(MetOH) ⎠ ⎛ 3 × fc(PropDi,Ref,HX1) ⎞ +⎜ ⎟ MW(PropDi) ⎠ ⎝ = (1 − fracCgas) ×

fc(Glycerol,Furnance,Ref) × 3 MW(Glycerol) (47)

Atomic balances to H, C, and O must also hold. Finally, we assume isothermal operation of the reformer as follows Q react =


Q prod =

(48)


(43)

⎟

j = {CO2 , CH4 , C2H6}

(44) (50)

We assume a feed to the reformer consisting of 10% of glycerol in water, based on the results by Liu et al.29 and Shabaker et al.28, because the less amount of water the less energy required in the furnace as long as the process is feasible. Lower concentrations result in an excessive consumption of energy to heat up the aqueous mixture. 3.2. Clean Up. The traces of hydrocarbons generated in the reforming stage are withdrawn from the gas stream using a PSA system. The typical working conditions for PSA systems are low temperature (25 °C) and moderate pressure (4.5 bar) so that there is adsorption of the different components on the bed.30 Typically a bed of silica gel is the most appropriate for the removal of hydrocarbons. We assume that the PSA retains any hydrocarbon left in the gas stream as well as the ammonia. Thus, the removal efficiency is 100% for hydrocarbons. Due to the low temperature, more water condenses in HX6, and it is discharged to Snk5 in Figure 6. 3.3. Hydrogen Production/Composition Adjustment. Once the main contaminants are eliminated, the ratio between H2 and CO may need to be adjusted so that the feed to the FT-reactor is appropriate for the optimal production of the diesel fraction, or we can produce only hydrogen. In order to perform such an adjustment, apart from modifying the operating conditions at the reformer, three alternatives are considered. The first one is the use of water gas shift to reduce the amount of CO by producing more H2, or to produce only hydrogen. The second is a bypass. Finally, a hybrid membrane/ PSA system with a bed of zeolite 13X is proposed to remove the excess of hydrogen, see Figure 7. It is possible to sell this

(45)

xCO2

∀j ∈ {H 2 , CO2 , CH4 , C2H6}

fracCgas = 0.86 + 3 × 10−3(T(Furnace,Ref) − 225)

mC

j = {CO2 , CH4 , C2H6}

;

(41)

Q prod − Q react ≠ 0 fc(j,Ref,HX1) = fracj ×

(39)

⎛ ⎞ B Pressure × 760 = exp⎜A − ⎟ (C + T(Furnace,Ref)) ⎠ ⎝

∑ ΔHjf |T(Reforming) j

6

(49) 7736



Article

packed bed. We assume that there is a 10% loss in the previous PSA system. The compression is modeled assuming polytropic behavior to determine the final temperature and the energy required. As a result of the cooling, water condenses in HX3. The amount of water condensed is determined by the saturation conditions of the exiting gas. In this PSA, it is assumed that only hydrogen is eliminated from the stream with an efficiency of 100%. The other gases pass though. Finally, all the streams mix adiabatically. The H2 to CO ratio is determined by the needs of the reactor to maximize the FT-diesel production. 3.4. CO2 Removal by PSA System. In this case, the syngas is used for the production of liquid fuels, and CO2 must be removed from the gas stream. Thus, the gas is first treated in a PSA system to remove CO2 by using zeolite 5A capable of removing 95% of CO2 in the stream.34,35 The absorption capacity is around 0.1 kg of CO2 per kg of zeolite. The system is modeled as two beds, one operating and the second one in regeneration, to allow continuous operation of the plant. The operating conditions are 25 °C and 4.5 bar; thus, in HX4, water condenses. Figure 8 shows a scheme of the process.

Figure 6. Details of the flowchart for the final hydrocarbon removal.

Figure 7. Superstructure flowchart for syngas composition adjustment.

surplus of hydrogen to increase the profitability of the process.31,32 Water Gas Shift. The reaction taking place in the water shift reactor is widely known CO + H 2O ↔ CO2 + H 2

Figure 8. Details of the flowchart for the PSA system for the removal of CO2.

3.5. Liquid Fuel Synthesis. The catalysts, either cobalt- or iron-based, the operating conditions at the FT-reactor, either low (200−240 °C) or high (300−350 °C) temperatures and pressures from 10 to 40 bar, determine the products and their distribution. In particular, the iron catalyst is the preferred choice because it provides high selectivity toward synthetic diesel, C10−C18. Moreover, the reactions with iron catalyst are usually conducted at 30 bar. The Fischer−Tropsch synthesis requires careful control of the H2:CO ratio to satisfy the stoichiometry of the synthesis reactions as well as to avoid deposition of carbon on the catalysts (coking). An optimal H2:CO ratio from 1:1 to 2:1 for the production of diesel and gasoline is recommended,24 while a minimum ratio of 1.7 is required for iron catalysts.36 Figure 9 shows the details of the flowchart for the synthesis where heat exchanger 7 heats the feed to the FT-reactor, Reactor 2. The reaction given in eq 53 is the desired one and is the most dominant reaction when applying cobalt-based FT-catalyst.

(51)

The conversion is calculated using a model developed from the experimental data by Choi et al.32 as a function of the molar ratio of water to CO (H2OtoCO) and the operating temperature (eq 52) proposed by Martı ́n and Grossman.27 CO_shift_conv (0.0044 × T(Reactor1) + 0.0924) × H 2OtoCO = 46815 H 2OtoCO + 2

(

T(Reactor1)

)

(52)

The products of the reactor are calculated as a function of the conversion in the reactor and the stoichiometry given by eq 51. The energy involved in the reaction is given by the heat of reaction (dH_shift_reac) and the conversion reached in the reactor. Bypass. It may be possible that the stream does not need any adjustment in the H2 to CO ratio for the optimal operation of the FT-reactor. H2 Membrane/PSA System. The stream to be treated in the membrane/PSA system for the recovery of pure hydrogen33 is cooled to 25 °C and compressed to 4.5 bar before entering the

⎛ m⎞ nCO + ⎜n + ⎟H 2 → CnHm + nH 2O ⎝ 2⎠ CO + 2H 2 → −CH 2 − +H 2O; ΔHFT = −165 kJ/mol (53) 7737



Article

⎞ ⎛ ⎞ ⎛ y CO ⎟ ⎜ ⎟ ⎜ α = 0.2332 × ⎜ ⎟ + 0.633⎟ ⎜ y y + ⎝ H2 CO ⎠ ⎠ ⎝ × (1 − 0.0039 × ((T_Synthesis + 273) − 533)) (56)

The mass balances in the reactor are simplified using the values calculated for the different fractions of the products from eq 55 and the total mass12 to determine the share of the different fractions, C1−C2, C3−C4, gasoline, diesel, and heavy product. 3.6. Separation and Hydrotreatment. The separation of hydrocarbons is very common in the petrochemical industry. The most important feature is that we are dealing with mixtures instead of single compounds. The gas fraction can be separated using a flash. However, in order to reduce the losses of the liquid products, the flash operates at the same pressure of the reactor, but we cool the mixture to 30 °C in HX8. The recovery of the flash is considered to be 100% for the gases and light hydrocarbons and 0% for the gasoline, diesel, wax, and water. The gases can be further separated into methane and butane fractions or can be used together as flue gas to produce energy. Next, the water is separated from the organic phase using a decanter. Subsequently, the three fuel fractions are separated as in any crude distillation system as shown in Figure 10, where the feed enters at 220 °C. According to Speight,41 the typical distillation towers for crude oil have 30 trays, and the typical temperatures are 125 °C for the top (gasoline), 220 °C for the diesel, and 280 °C at the bottom for the heavier components. The outlet of the fractionation columns are mainly gasoline from the top, diesel from the middle, and the heavy fraction from the bottom. The reflux ratio is assumed to be 2.42 These kinds of columns do not have reboilers because steam is directly injected at the bottom. We assume that we recover the steam used and recycle it after heating it again, and thus, the energy consumption of the column is given by the furnace to heat the feed to the operating temperature, around 220 °C (HX11 in Figure 10), and the steam required is assumed to be around 0.18 kg of steam per kg of residue, based on the results by Jonas and Pujado.43 The conditions are on the conservative side due to the range of results reported in the literature and

Figure 9. Details of the flowchart for the FT-reactor.

Apart from this main reaction, a number of other chemical reactions also take place such as methanation, the first reaction in eq 54, the water gas shift (WGS), the second reaction, and the Boudouard equilibrium. In particular, when using an ironbased (Fe) catalyst, the WGS, second reaction in eq 54, the reaction also readily takes place because Fe catalyzes the WGS enabling the operation at a lower temperature. On the other hand, the methanation reaction and the Boudouard reaction are undesirable. CO + 3H 2 → CH4 + H 2O; CO + H 2O ↔ CO2 + H 2 ; 2CO ↔ C + CO2 ;

ΔH 298 = 247 kJ/mol ΔH 298 = −41 kJ/mol

ΔH 298 = −172 kJ/mol (54)

In order to model the length of the hydrocarbons obtained, we use the Anderson−Schulz−Flory (ASF) distribution that describes the probability of hydrocarbon chain growth.37 The probability of chain growth can be denoted as α. The operating temperature of the reactor and the inlet H2 to CO ratio are calculated for the optimal production of diesel based on the correlation obtained by Hyun−Seob38 for α (eq 56) combined with the model for the mass fraction of each hydrocarbon (wi, i = number of C) that assumes that the Fischer−Tropsch reactor works as a polymerization reactor39 (eq 55). A conversion of 0.8−0.9 in CO is considered.40 wi = α i − 1(1 − α)2 × i

(55)

Figure 10. Details of the flowchart for the diesel and gasoline fractionation. 7738



Article

the fact that most of the data are reported for crude and not for FT products.42−45 The bottoms of the column can be upgraded to obtain more gasoline and diesel. Hydrocracking (Reactor 3) is the best option in order to produce diesel virtually free from aromatics.36 The temperature of operation is given by the optimum performance toward the production of diesel; thus, it is a decision variable. To model the hydrocracking reactor, Reactor 3, we use the experimental data for the conversion and selectively as a function of the temperature from the paper presented by Bezergianni et al.26 We correlate these profiles12 using polynomials to develop a reduced order model for the hydrocracking reactor (eqs 57−58). −4

XR3 = 1.85714 × 10

ZSR = C H2 × fc(H 2,Spl2,Snk3) − C NatGas × Q(Furnance) − CSteam × (Q(HX2)/λ + fc(H 2O,Src4,Reactor1) + fc(H 2O,Src2,Furnance))

ZAR = C H2 × fc(H 2,Spl2,Snk3) − C NatGas × Q(Furnance) − CO2 × fc(O2 ,Src3,Furnance) − CSteam × (Q(HX2)/λ + fc(H 2O,Src4,Reactor1) + fc(H 2O,Src2,Furnance))

× (T_reactor3) − 0.128829 (57)

ZAR = CH2(fc(H 2,Spl2,Snk3) + fc(H 2,MS4,Snk4)) + C Diesel fc(Diesel,HX11,Col1) + −CSteam

Sdiesel = −1.4286 × 10−4 × (T_reactor3)2 + 9.9514 × 10−2 × T_reactor3 − 16.383

× (Q(HX7)/λ + fc(H 2O,Src4,Reactor1)

(58)

+ fc(H 2O,Src2,Furnance)) − C NatgasQ(Furnance)

The reactor also requires hydrogen. We assume 0.1 N m3 per liter of liquid fuels to be left with the products. This amount represents around 7.5% of the total hydrogen fed to the hydrocracker.41 These values depend on the composition of the feed and the catalysts and operating conditions, and a range is available in the literature. Thus, the products from the reactor are calculated based on the conversion and selectivity. The products will be separated by cooling them to 30 °C in a flash so that the gasoline and diesel fractions are recycled to the column to be separated, while the H2 is recycled to the system so that only a small amount of makeup of hydrogen is fed, mainly that which remains with the liquid products. Meanwhile, the liquids (gasoline, diesel, and heavy) are recycled to the column to be separated again. 3.7. Solution Procedure. The original MINLP problem is decomposed into 3 NLP subproblems, one for each reforming mode. Because APR produces CO2 and H2 directly, we first evaluate the production of hydrogen using the three alternatives. Next, for autoreforming and steam reforming, we evaluate the possibility of producing FT-diesel, hydrogen, or both at the same time by solving the superstructure given in Figure 2 but without considering APR. Because the price for gasoline is higher than that of diesel, to optimize the production of FT-diesel, we do not consider gasoline in the objective function. Each of the subproblems, one per reforming mode, is solved as an NLP to optimize the operating conditions of the reformer, the WGSR, the Fischer−Tropsch reactor, and the hydrocracking unit. The objective function to be maximized is the profit involving the income from the production of diesel and the costs for use of energy to prepare the feed for the reformer, the WGSR and the FT-reactor, as well as the hydrogen used in hydrocracking. We do not include investment costs based on the fact that the flowchart actually does not change from one reforming mode to the other. Equations 59−61 are used in the optimization of the hydrogen production for the three reforming modes, APR, SR, and AR, respectively

− CO2 fc(O2 ,Src1/2,Furnance) − C H2 0.01 × fc(Nafta,HX11,11Col1)

(62)

ZSR = C H2 × (fc(H 2,Spl2,Snk3) + fc(H 2,MS4,Snk4)) + C Diesel × fc(Diesel,HX11,Col1) − CSteam × (Q(HX7)/λ + fc(H 2O,Src4,Reactor1) + fc(H 2O,Src2,Furnace)) − C NatGas × Q(Furnace) − C H2 × 0.01 × fc(Nafta,HX11,Col1)

(63)

Thus, the main decision variables are the steam and/or oxygen added to the reformer and its operating temperature, the split fraction at the water gas shift reactor and hydrogen PSA system for the composition adjustment, the water gas shift operating conditions (temperature and steam needed), the operating conditions at the Fischer−Tropsch reactor (temperature and CO to H2 ratio), and the hydrocracking conditions.

4. RESULTS AND DISCUSSION The size of the plant is fixed to process 1 kg/s of glycerol, which is approximately the production of glycerol in typical biodiesel production facilities.3 The cost for utilities is updated from the literature ($0.019/kg, steam, $0.057/t cooling water,46 $0.06/kWh,47 and $0.021/kg oxygen48). The cost of hydrogen is taken to be $1.6/kg as established by the U.S. Department of Energy (USDOE) for the long term. We consider $1/kg of liquid fuel produced, and the cost of natural gas is $4.687/ million BTU ($4.44/GJ). The generation of an excess of steam is considered as a revenue of 0.0077$/kgsteam (updated from Smith and Varbanov49). We determine the production cost and the investment based on Coulson’s methodology.50 4.1. Hydrogen Production. In the first case, we consider the production of hydrogen alone using the three reforming technologies. Table 1 summarizes the optimal operating conditions for the main variables. Actually, autoreforming and steam reforming use almost the same amount of steam, but due to the fact that autoreforming burns part of the glycerol to obtain the energy for the steam reforming reactions, the

ZAPR = C H2 × fc(H 2,Ref,Sep1) − C NatGas × Q(Furnace) − C Water × fc(H 2O,Src1,Furnace)

(61)

For the simultaneous production of hydrogen and liquid fluids, we use eqs 62 and 63 for autoreforming and steam reforming, respectively

2

× T_reactor3 + 22.6931

(60)

(59) 7739



Article

4.2. Simultaneous Liquid Fuels and Hydrogen Production. The simultaneous production of hydrogen and liquid fuels from glycerol means solving the complete superstructure deciding on the reforming process, between SR and AR, on the separation of hydrogen and the technology used, as well as on the syngas composition adjustment, ratio H2 to CO, for the synthesis of liquid fuels following FT reactions. According to Wang et al.,24 increasing the inlet H2 to CO ratio causes an increase in lighter distillates and a decrease in heavier distillates. Higher gasoline and diesel product selectivity is obtained when the inlet H2 to CO ratio is between 1 and 2. However, Dry36 reported that for iron catalysts, a minimum of 1.7 is required. Adjusting the composition requires energy, and thus, the results of the optimization lead to the lower bound 1.7. Table 3 presents the main operating parameters of both

Table 1. Process Operating Conditions for Hydrogen Production

oxygen added (mol/mol) steam added (mol/mol) temperature (°C)

autoreforming

steam reforming

aqueous phase

0.187 3.038 746

2.738 587

265

Table 3. Process Operating Conditions for Hydrogen and Liquid Fuels Production from Glycerol oxygen added (mol/mol) steam added (mol/mol) reforming temperature (°C) FT temperature (°C) α molar ratio H2/CO

Figure 11. Net energy balance for the production of hydrogen (1 kg/s of glycerol).

operating temperature of this option is higher. In the case of the aqueous phase reforming, the operating temperature is the lowest of the three so as to operate in liquid phase. Figure 11 presents the cooling and energy (thermal and electrical) needs for each of the processes. The APR generates energy in the form of low pressure steam but requires a large amount of cooling due to the fact that it uses liquid water at high pressure to decompose the glycerol. Either steam reforming or autoreforming are energy intensive due to glycerol gasification. Next, after heat integration of the streams using SYNHEAT,51 we perform an economic analysis. The correlations for the cost of the main units can be found in Martı ́n and Grossmann.2 The results are shown in Table 2 assuming a

autoreforming

steam reforming

0.101 1 684 200 0.9 1.7

NA 1 605 201 0.89 1.7

reforming alternatives. We can see that the operating conditions at the FT-reactor lead to the highest fraction of diesel substitutes with α equal to 0.9 by tuning the reaction temperature and the molar H2 to CO ratio. Figure 13 shows

Table 2. Hydrogen Production (glycerol cost $0/kg) utilities contribution (%) biofuel yield (kg/kg) production cost ($/kg) investment ($ MM)

autoreforming

steam reforming

APR

36.0 0.123 0.83 14.8

33.7 0.084 0.87 10.6

6.4 0.111 0.55 11.4

glycerol cost of $0/gal ($0/L). Later in the paper, we evaluate the effect of the cost of glycerol on the production of hydrogen. Even though the highest yield to hydrogen corresponds to autoreforming, the higher consumption of utilities, oxygen and steam, and the larger number of process units for gas purification result in the fact that aqueous phase reforming is the most attractive process. The production cost of the APR based process is $0.55/kg, at zero cost of glycerol, with the lowest investment cost of $11.4 MM (Figure 12).

Figure 13. Optimized distribution of products (a) autoreforming and (b) steam reforming.

the optimal product distribution targeting FT-diesel production. The results of the optimization indicate that the optimal operation of the process produces hydrogen and liquid fuels simultaneously so that the yield from glycerol to fuels is maximized.

Figure 12. Optimal flowchart for the production of hydrogen from glycerol. 7740



Article

product is FT-diesel, but the share in mass of gasoline and hydrogen produced is similar. There is a trade-off between the investment cost and the production cost because autoreforming has higher yield and thus lower production cost, but the investment is slightly higher. The main advantage of autoreforming is the fact that it produces a surplus of hydrogen (we assume $1.58/kg USDOE final target) that does not only cover the needs of hydrocracking but also provides extra revenue by increasing the total yield of fuels and reducing the utilities cost, even though the selectivity to diesel is slightly lower than using steam reforming. The optimal flowchart is shown in Figure 15 including the major operating conditions. The optimal operating conditions at the Fischer−Tropsch reactor look for maximum yield to FT-diesel. In this way, the ratio for H2 to CO is maintained at the lower bound indicating that this variable is expensive to adjust. Actually the importance of this variable is shown in the marginal value reported at the optimal solution. It is interesting to note that if we use a H2 to CO ratio of 2, widely used in industry, the other operating conditions at the reactor remain the same. However, the objective function is then reduced by 11%, the steam consumed in the reformer increases by 25%, and the yields to diesel and hydrogen are reduced by 7.5% and 11%, respectively, while that to green gasoline increases by 8.5%. If we compare the production of hydrogen alone with the simultaneous production of liquid fuels and hydrogen, the first thing that stands out is the higher efficiency to fuels in the latter option. The yield to fuels is more than twice taking advantage of a higher fraction of the raw material. However, that comes at a high expense. The investment is at least 50% higher. On the other hand, the production costs of the process that produces liquid fuels and hydrogen are more competitive because we can also obtain credit for the hydrogen obtained. 4.3. Raw Materials Cost Sensitivity Analysis. 4.3.1. Hydrogen Production. The cost of glycerol is an uncertain parameter. The increase in its production capacity parallel to biodiesel production is expected to result in a decrease in its price. The increase in the price of glycerol does not affect the optimal operating conditions at the plant. However, it has an impact on the production costs. Ahmed and Papadias6

Figure 14. Net energy balance for the production of hydrogen and liquid fuels (1 kg/s of glycerol).

In Figure 14, we present the net energy balance after heat integration and heat exchanger network design for both reforming technologies for the production of hydrogen and liquid fuels. Autoreforming requires less energy and more cooling than steam reforming because the process is a combination of exothermic and endothermic reactions, while steam reforming involves only endothermic reactions resulting in lower final temperatures and higher energy inputs for the reforming to take place. Next, we perform an economic evaluation for both alternatives. The results are summarized in Table 4, assuming Table 4. Liquid Fuels Production (glycerol cost $0/kg) biofuel yield (kg/kg) selectivity to diesel selectivity to H2 production cost ($/gal) [$/L] investment ($ MM)

autoreforming

steam reforming

0.23 0.58 0.20 0.02 [0.005] 18.2

0.16 0.64 0.18 0.73 [0.19] 15.6

a glycerol cost of $0/kg. In the next section, we evaluate the effect of the glycerol price on the cost of liquid fuels. The main

Figure 15. Optimal flowchart for the simultaneous production of hydrogen and liquid fuels. 7741



Article

($0.10/L), merely $0.04/gal ($0.01/L) more than the base case facility that sells the glycerol to the market at $0.3/kg. 4.3.2. Liquid Fuels and Hydrogen Production. As in the case when only hydrogen is produced, the cost of glycerol does not affect the operating conditions nor the preferred reforming mode but only the production cost. If the glycerol cost reaches $0.102/lb6, the production cost of liquid fuels following the optimal alternative that uses autoreforming increases up to $3.70/gal ($0.98/L), while to meet the $1/gal ($0.26/L) of liquid fuel, the price of glycerol has to be no higher than $0.025/lb ($0.055/kg), which is half the value expected in the long term by the USDOE (Figure 17). If we compare the use of

considered a glycerol cost of $0.102/lb ($0.225/kg), which results in a production cost of hydrogen of $2.8/kg, which is reasonable for the current price but almost twice as expensive as the expected cost of $1.58/kg in the long run by the USDOE. To reach this goal, the glycerol cost must be $0.052/lb ($0.114/kg), which in fact is the expected final value for glycerol according to the USDOE.52 Figure 16 presents the profile of the cost of hydrogen as a function of the glycerol cost.

Figure 16. Competition between the use of switchgrass and glycerol for the production of hydrogen.

Figure 17. Competition between the use of switchgrass and glycerol for the production of liquid fuels.

On the other hand, if we compare the production of hydrogen from glycerol with that which uses switchgrass as raw material27 (Figure 16) the hydrogen from glycerol is competitive if the cost of the switchgrass reaches $90/t. We acknowledge that the production capacity of both plants is not the same. The lignocellulosicbased facility produces 60 × 106kg/yr, while the process models in this paper only reach 3.45 × 10 × 106kg/yr. However, what is similar is the production capacity of the biodiesel plant, which generates the glycerol as byproduct and the hydrogen plant. In this sense, this comparison is more realistic than if we scale up the production plant that uses glycerol as raw material to the levels of a biorefinery of switchgrass because we will probably not be able to obtain that amount of glycerol within a reasonable distance. As a result of the discovery of shale gas, the lowest value for natural gas price was reached around $2/MMBTU ($1.86/GJ). Thus, the production cost of hydrogen using this value is reduced by $0.2/kg, so that the use of glycerol is competitive with switchgrass if it reaches $80/t. Furthermore, it is possible to reach the target for the hydrogen price given by the USDOE for glycerol costs of $0.06/lb ($0.13/kg). Finally, the CO2 produced together with the hydrogen can be recycled and used to produce biodiesel and with it more glycerol, closing the cycle. In this way, there will be no emissions of CO2 from the process. In fact, if we integrate this plant that transforms glycerol into hydrogen with a biodiesel facility from algae that simultaneously produces bioethanol and biodiesel based on enzymatic transesterification,53 the liquid fuels capacity remains the same 90 MMgal/yr (340 × 106 L/yr), the investment costs increase by 5.7% up to $191 MM, and the production costs increases up to $0.45/gal ($0.12/L) before considering H2 as a credit. However, if we evaluate the profit of selling the hydrogen, assuming $1.58/kg of hydrogen, the production costs of the liquid fuel is reduced to $0.39/gal

glycerol and switchgrass for the production of liquid fuels, we see that as in the previous case for the hydrogen production switchgrass must reach $90/t and glycerol must reduce its cost to less than $0.05/lb ($0.011/kg). In Figure 17, the flue gas from the FT-reactor has not been considered as a source of energy. However, there is potential to improve the profitability of the process by using the flue gas to generate steam and sell it. In doing so, we can obtain around 2.3 MW of energy from each of the technologies, either autoreforming or steam reforming, which represents a reduction of $0.35/gal ($0.09/L) of biofuel. In this case the production cost of biofuels using the autoreforming path is $3.24/gal ($0.86/L) if the glycerol reaches $0.102/lb ($0.225/kg)6, and for the limit of $1/gal ($0.26/L), the price of glycerol can go up to $0.04/lb ($0.088/kg). Furthermore, the price of natural gas is volatile with the development of shale gas. If we consider the lowest value of $2/MMBTU ($1.86/GJ), together with the use of flue gas as a revenue, the production cost of liquid fuels is reduced around $0.5/gal ($0.13/L) with respect to the values presented by the red line in Figure 17. In this case, the production of liquid fuels out of glycerol becomes competitive if the switchgrass reaches $80/t and/ or the glycerol price is $0.05/lb ($0.110/kg), which is the USDOE target. Finally, we can integrate this process with the production facility that is capable of producing bioethanol and biodiesels simultaneously from algae using enzymatic transesterification.53 That being the case, the liquid fuels production increases by only 1.5% up to 91.4 MMgal/yr (346 ×106 L/yr). However, the investment cost of such an integrated facility adds up to $195 MM, 7.5% higher than the facility that has glycerol as byproduct, and the production cost increases by $0.075/gal ($0.02/L) to $0.425/gal ($0.11/L) with respect 7742



Article

fracCgas: = fraction of C in the gas phase H2OtoCO = molar ratio H2O and CO mC = mass of carbon (kg) Mass product = mass flow of liquid products (kg/s) MW(i) = molecular weight of component i (kg/kmol) ni = mol of component i (i) i = {H2,CO2, CO, CH4, H2O, C2H4, C2H2, C,O2} (kmol/s) Oxygen_add = mol oxygen per mol of carbon Pressure = operation pressure (bar) Q(j) = heat exchanged in unit j (kW) Q_prod = energy products (kJ/s) Q_reac = energy reactants (kJ/s) steam_add = mol steam per mol of carbon S_diesel = selectivity to diesel S_gasoline = selectivity to gasoline T_amb = ambient temperature (20 °C) T_synthesis = temperature at FT-reactor (°C) T_reactor3 = temperature at hydrocracking (°C) T_reforming = average temperature at the steam reformer or autoreformer (°C) T(unit1,unit2) = temperature of the stream from unit 1 to unit 2 (°C) Tb = boiling temperature (°C) Waterdecomp = mass of water decomposed at gasifier (kg/s) XR3 = conversion at hydrocracker x(J,unit1,unit2) = mass fraction of stream from unit 1 to unit 2 xi = mass fraction of species i yi = molar fraction of species i λ = vaporization heat (kJ/kg) Z = objective value ($/s)

to the base case considering the credit by the hydrogen produced.

5. CONCLUSIONS A superstructure model has been formulated for the simultaneous production of hydrogen and liquid fuels including three glycerol reforming modes, gas clean up, several alternatives for syngas composition adjustment, sour gases removal, and FT-synthesis for liquid fuels production. The corresponding MINLP problem is solved by decomposing it into hydrogen production and simultaneous production of hydrogen and liquid biofuels production because aqueous phase reforming can only produce hydrogen. Each subproblem is optimized to determine the operating conditions of the main units. The results indicate that the production of liquid fuels from glycerol is not competitive with the use of lignocellulosic switchgrass unless the price for glycerol is very low. For the production of liquid fuels, AR is recommended, but from an economic point of view, it is not competitive with the use of switchgrass as raw material. However, for the production of hydrogen alone, aqueous phase reforming is recommended, and the process becomes competitive if the price of glycerol drops below $0.05/lb ($0.110/kg), around one-half the prediction of the USDOE in the medium term. Hydrogen production from glycerol produces CO2 that can be recycled to the algae ponds that produce the oil that is used for biodiesel and glycerol production. Further experimental results at the pilot plant level are needed to validate these results. The integration of this process with an algae-based biodiesel plant results in promising economic values, above all in the case of producing hydrogen from glycerol with a moderate increase in investment and production costs but including hydrogen as a product of the biorefinery.

■

Subindexes

AR = autoreforming APR = aqueous phase reforming C = carbon C2H2 = ethylene C2H6 = ethane C2H4 = ethene CO = carbon monoxide CO2 = carbon dioxide EtOH = ethanol H2 = hydrogen H2O = water MetOH = methanol NatGas = natural gas O2 = oxygen PropDi = propanediol SR = steam reforming

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. Notes

The authors declare no competing financial interest.

■

ACKNOWLEDGMENTS The authors gratefully acknowledge the NSF Grant CBET0966524 and the Center for Advanced Process Decision-making at Carnegie Mellon University. Dr. Mariano Martiń acknowledges Salamanca Research for software licenses.

■

NOMENCLATURE α = chain length A,B,C = Antoine coefficients Ci = cost $/kg of species i CO_shift_conv = conversion of CO in the water shift reactor COtoH2 = molar ratio CO and H2 at mix 1 Condensed_j = water condensed before equipment j (kg/s) Cp (J) = liquid phase heat capacity of element J (kJ/kg K) cp_v(J) = vapor phase heat capacity of element J (kJ/kg K) dH_shift_reac = heat of reaction for the water gas shift (kJ/kmol) fc(j,unit1,unit2) = individual mass flow rate (kg/s) of component j from unit 1 to unit 2 F(unit1,unit2) = total mass flow rate (kg/s) from unit 1 to unit 2

Units

■

Compres: = compressor Flash = flash unit Fur/Furnace = furnace HX: = heat exchanger MS: = molecular sieve ref: = reforming furnace Sep: = phase separator Src: = sources Snk = sinks

REFERENCES

(1) Karuppiah, R.; Peschel, A.; Grossmann, I. E.; Martín, M.; Martinson, W.; Zullo, L. Energy optimization of an ethanol plant. AIChE J. 2008, 54 (6), 1499−1525.

7743



Article

(2) Martín, M.; Grossmann, I. E. Energy optimization of bioethanol production via gasification of switchgrass. AIChE J. 2011, 57 (12), 3408−3428. (3) Martín, M.; Grossmann, I. E. Simultaneous optimization and heat integration for biodiesel production from cooking oil and algae. Ind. Eng. Chem. Res. 2012, 51 (23), 7998−8014. (4) Pagliaro, M.; Rossi, M. Future of Glycerol, 2nd ed; Royal Society of Chemistry: Cambridge, U.K., 2010. (5) Fan, X.; Burton, R.; Zhou, Y. Glycerol (byproduct of biodiesel production) as a source for fuels and chemicals − Mini review. Open Fuels Energy Sci. J. 2010, 3, 17−22. (6) Ahmed, S.; Papadias, D. Hydrogen from Glycerol: A Feasibility Study. Presented at the 2010 Hydrogen Program Annual Merit Review Meeting, June 8, 2010, Washington DC. http://www.hydrogen.energy. gov/pdfs/review10/pd003_ahmed_2010_o_web.pdf; http://www. hydrogen.energy.gov/pdfs/progress10/ii_a_3_ahmed.pdf. (7) Wilhelm, D. J.; Simbeck, D. R.; Karp, A. D.; Dickenson, R. L. Syngas production for gas-to-liquids applications: technologies, issues and outlook. Fuel Process. Technol. 2001, 71, 139−148. (8) Cecil, W. On the Application of Hydrogen Gas To Produce a Moving Power in Machinery, 1820. http://www.eng.cam.ac.uk/ DesignOffice/projects/cecil/cecil.pdf (accessed July 2012). (9) Cole, D. E. Issues Facing the Auto Industry: Alternative Fuels, Technologies, and Policies. ACP Meeting, Eagle Crest Conference Center, Ypsilanti, Michigan, June 20, 2007. (10) Rand, D. A. J.; Dell, R. M. Hydrogen Energy: Challenges and Prospects; The Royal Society of Chemistry: Cambridge, U.K., 2008. (11) Swanson, R. M.; Platon, A.; Satrio, J. A.; Brown, R. C. Technoeconomic analysis of biomass-to-liquids production based on gasification. Fuel 2010, 89 (Suppl.1), S11−S19. (12) Martín, M.; Grossmann, I. E. Process optimization of FT-diesel production from lignocellulosic switchgrass. Ind. Eng. Chem. Res. 2011, 50 (23), 13485−13499. (13) Adhikari, S.; Fernando, S.; Gwaltney, S. R.; Filip To, S. D.; Bricka, R. M.; Steele, P. H.; Haryanto, A. A thermodynamic analysis of hydrogen production by steam reforming of glycerol. Int. J. Hydrogen Energy 2007, 32, 2875−2880. (14) Douette, A. M. D.; Turn, S. Q.; Wang, W.; Keffer, V. I. Experimental investigation of hydrogen production from glycerin reforming. Energy Fuels 2007, 21, 3499−3504. (15) Wang, X.; Li, S.; Wang, H.; Liu, B.; Ma, X. Thermodynamic analysis of glycerin steam reforming. Energy Fuels 2008, 22, 4285− 4291. (16) Da Silva, A. L.; Malfatti, C. F.; Müller, I. L. Thermodynamic analysis of ethanol steam reforming using Gibbs energy minimzation method: A detailed study of the condtions of carbon deposition. Int. J. Hydrogen Energy 2009, 34, 4321−4330. (17) Feng, X.; Wang, L.; Min, S. Industrial emergy evaluation for hydrogen production system from biomass and natural gas. Appl. Energy 2009, 86, 1767−1773. (18) Gao, N.; Li, A.; Quan, C. A novel reforming method for hydrogen production from biomass steam gasification. Bioresour. Technol. 2009, 100, 4271−4277. (19) Ji, P.; Feng, W.; Chen, B. Production of ultrapure hydrogen from biomass gasification with air. Chem. Eng. Sci. 2009, 64, 582−592. (20) Ji, P.; Feng, W.; Chen, B. Comprehensive simulation of an intensified process for H2 production from steam gasification of biomass. Ind. Eng. Chem. Res. 2009, 48, 3909−3920. (21) Kunkes, E. L.; Soares, R. R.; Simonett, D. A.; Dumesic, J. A. An integrated catalytic approach for the production of hydrogen by glycerol reforming coupled with water-gas shift. Appl. Catal., B 2009, 90, 693−698. (22) Mueller-Langer, F.; Tzimas, E.; Kalschmitt, M.; Peteves, S. Techno-economic assessment of hydrogen production process for the hydrogen economy for the short and medium term. Int. J. Hydrogen Energy 2007, 32, 3797−3810. (23) Grossmann, I. E.; Caballero, J. A.; Yeomans, H. Mathematical programming approaches to the synthesis of chemical process systems. Korean J. Chem. Eng. 1999, 16, 407−426.

(24) Wang, Y.; Fan, W.; Liu, Y.; Zeng, Z.; Hao, X.; Chang, M.; Zhang, C.; Xu, Y.; Xiang, H.; Li, Y. Modeling of the Fischer−Tropsch synthesis in slurry bubble column reactors. Chem. Eng. Process. 2008, 47, 222−228. (25) Opdal, O.; Skreiberg, O. Production of Synthetic Biodiesel via Fischer−Tropsch Synthesis: Biomass-to-Liquids in Namdalen, Norway; Project Report; Department of Energy & Process Engineering, Faculty of Engineering Science and Technology, Norwegian University of Science and Technology: Trondheim, Norway, 2006. (26) Bezergianni, S.; Kalogianni, A.; Vasalos, I. A. Hydrocracking of vacuum gas oil-vegetable oil mixtures for biofuels production. Bioresour. Technol. 2009, 100, 3036−3042. (27) Martín, M.; Grossmann, I. E. Energy optimization of hydrogen production from biomass. Comput. Chem. Eng. 2011, 35 (9), 1798− 1806. (28) Shabaker, J. W.; Huber, G. W.; Dumesic, J. A. Aqueous-phase reforming of oxygenated hydrocarbons over Sn-modified Ni catalysts. J. Catal. 2004, 222, 180−191. (29) Liu, B.; Zhang, Y.; Tierney, J. W.; Wender, I. Hydrogen Generation from Glycerol via Aqueous Phase Reforming. Presentation at CFFS Annual Meeting, Daniels, WV, August 1, 2006. (30) Olofsson, I.; Nordin, A.; Söderlind, U. Initial Review and Evaluation of Process Technologies and Systems Suitable for CostEfficient Medium-Scale Gasification for Biomass to Liquid Fuels; ETPC Report 05-02; Umeå University: Umeå, Sweden, 2005. (31) Neves, C. F. C.; Schvartzman, M.M. A.M. Separaçao de CO2 per meio da tecnologiá PSA. Quim. Nova 2005, 28, 622−628. (32) Choi, D.; Chipman, D. C.; Bents, S. C.; Brown, R. C. A. Techno-economic analysis of polyhydro-xyalkanoate and hydrogen production from syngas fermentation of gasified biomass. Appl. Biochem. Biotechnol. 2010, 160 (4), 1032−1046. (33) IST World. http://www.ist-world.org. (34) Ko, D.; Siriwardane, R.; Biegler, L. T. Optimization of a pressure-swing adsorption process using Zeolite 13X for CO 2 sequestration. Ind. Eng. Chem. Res. 2003, 42, 339−348. (35) Reynolds, S.; Ebner, A.; Ritter, J. New pressure swing adsorption cycles for carbon dioxide sequestration. Adsorption 2005, 11 (Suppl. 1), 531−536. (36) Dry, M. E. The Fischer−Tropsch process: 1950−2000. Catal. Today. 2002, 71, 227−241. (37) Schulz, H.; Schaub, G.; Claeys, M.; Riedel, T. Transient initial kinetic regimes of Fischer−Tropsch synthesis. Appl. Catal. A 1999, 186, 215−227. (38) Hyun-Seob, S.; Ramkrishna, D.; Trinh, S.; Wright, H. Operating strategies for Fischer−Tropsch reactors: A model-directed study. Korean J. Chem. 2004, 21, 308−317. (39) Fürnsinn, S.; Ripfel, K.; Rauch, R.; Hofbauer, H. Diesel aus Holz - Die FS Synthese als zukunftsweisende Technologie zur Gewinnung flüssiger Brennstoffe aus Biomasse, 4. Internationale Energiewirtschaftstagung an der TU Wien, 2005. (40) Park, C. S.; Norbeck, J. M. Synthetic Diesel Production from Carbonaceous Feed Stocks. www.eri.ucr.edu/ISAFXVCD/ ISAFXVAF/SyDPCFS.pdf (accesed June 2012). (41) Speight, J. G. The Chemistry and Technology of Petroleum, 2nd ed.; Marcel Dekker, Inc.: New York, 1991. (42) Parkash, S. Refining Process Handbook. Gulf Professional Publishing, Elsevier: Houston, TX, 2003. (43) Jones, D. S. J.; Pujadó, P. R. Handbook of Petroleum Processing; Springer: Dordrecht, The Netherlands, 2006. (44) Marcel-Dekker Encyclopedia of Chemical Processing and Design; Marcel Dekker, Inc.: New York, 1997. (45) More, R. K.; Bulasara, V. K.; Uppaluri, R.; Banjara, V. R. Optimization of crude distillation system using Aspen Plus: Effect of binary feed selection on grass-root design. Chem. Eng. Res. Des. 2010, 88, 121−134. (46) Franceschin, G.; Zamboni, A.; Bezzo, F.; Bertucco, A. Ethanol from corn: A technical and economical assessment based on different scenarios. Chem. Eng. Res. Des. 2008, 86 (5), 488−498. 7744



Article

(47) Balat, M.; Balat, H.; Ö z, C. Progress in bioethanol processing. Prog. Energy Combust. Sci. 2008, 34 (5), 551−573. (48) Forsberg, C. W.; Gorensek, M. B. Relative Economic Incentives for Hydrogen from Nuclear, Renewable, and Fossil Energy Sources, American Institute of Chemical Engineers, Annual Meeting, Salt Lake City, UT, November 4−9, 2007. (49) Smith, R.; Varbanov, P. What’s the price of steam? Chem. Eng. Prog. 2005, 29−33. (50) Sinnott, R. K. Coulson & Richardson’s Chemical Engineering, 3rd ed.; Butterworth Heinemann: Burlington, MA, 1999. (51) Yee, T. F.; Grossmann, I. E. Simultaneous optimization models for heat integration. II. Heat exchanger networks synthesis. Comput. Chem. Eng. 1990, 28, 1165−1184. (52) Chi, Z.; Pyle, D.; Wen, Z.; Frear, C.; Chen, S. A laboratory study of producing docosahexaenoic acid from biodiesel-waste glycerol by microalgal fermentation. Process Biochem. 2007, 42, 1537−1545. (53) Martín, M.; Grossmann, I. E. Optimal engineered algae composition for the integrated simultaneous production of bioethanol and biodiesel. AIChE J. 2013, 59 (8), 2872−2883.

7745


Optimal Simultaneous Production of Hydrogen and Liquid Fuels from

Recommend Documents