Energy & Fuels 2007, 21, 3499–3504
3499
Experimental Investigation of Hydrogen Production from Glycerin Reforming Aurelien M. D. Douette, Scott Q. Turn,* Wuyin Wang, and Vheissu I. Keffer Hawaii Natural Energy Institute, UniVersity of Hawaii, 1680 East West Road, POST 109, Honolulu, Hawaii 96822 ReceiVed August 14, 2006. ReVised Manuscript ReceiVed July 6, 2007
A series of tests were performed to investigate reforming of reagent-grade propane-1,2,3-triol, (C3H8O3) commonly called glycerin, to produce a H2 rich gas. Effects of the operating parameters, oxygen to carbon ratio, steam to carbon ratio, and temperature, were determined using a factorial experimental design. A mathematical model defining the effect of the three parameters was derived and used to improve the hydrogen yield. From the range of experimental conditions, it was concluded that the oxygen to carbon ratio, as well as the interaction between oxygen to carbon ratio and temperature had the most important effects on H2 yield. A 4.5 mol quantity of hydrogen was produced per mole of glycerin at experimental conditions of oxygen to carbon ratio of 0, steam to carbon ratio of 2.2, and temperature of 804 °C. This is 65% of the maximum theoretical H2 yield, and 90% of the H2 yield predicted by thermochemical equilibrium. A 1.4 mol quantity of CO was also produced per mole of glycerin, presenting the potential for additional production of 1.4 mol H2/mol glycerin. A water gas shift reactor was added to the process and operated at 369 °C, producing a final yield of 5.3 mol H2/mol glycerin, 75% of the maximum stoichiometric hydrogen yield. Crude glycerin obtained from biodiesel production was finally tested (without the water gas shift reactor) as a feed and compared with reagent-grade glycerin results. The initial crude glycerin hydrogen yield, 4.4 mol H2/mol glycerin, was almost identical to that of reagent-grade glycerin, but carbon formation and coking increased the pressure drop through the catalyst bed causing the test to be terminated. Possible contaminants, chloride and sodium cations, present in crude glycerin as byproducts of biodiesel synthesis were added to reagent-grade glycerin and tested in the reformer, producing results similar to those observed for the crude glycerin reforming test.
Introduction Society is increasingly looking for clean and renewable fuels to offset the negative effects of fossil fuel use including greenhouse gas emissions and consumption of limited resources. Two possible renewable energy supplies are biodiesel and hydrogen. The U.S. has a biodiesel production capacity of about 350 million gallons with a 2005 production of 100 million gallons.1 Ten million gallons of crude glycerin (C3H8O3, IUPAC name propane-1,2,3-triol) were also produced as byproduct. Biodiesel has become more competitive against petroleum diesel due to the higher prices of crude oil and increased demand for environmentally acceptable fuels. It is anticipated that biodiesel and crude glycerin production will continue to grow. Hydrogen is a clean energy source with uses including ammonia production, petroleum processing, and power generation in fuel cells.2–6 Fossil fuels are currently used as the feed stock to satisfy 95% of U.S. hydrogen demand, and steam reforming of natural gas * Corresponding author. E-mail:
[email protected]. (1) Butzen, S. Biodiesel production in the U.S. Pioneer, A Dupont Company. http://www.pioneer.com/CMRoot/Pioneer/media_room/biofuels/ documents/biodiesel.pdf (accessed Aug 2006). (2) Ramachandran, R.; Menon, R. K. Int. J. Hydrogen Energy 1998, 23, 593–598. (3) Ancheyta, J.; Rana, M. S.; Furimsky, E. Catal. Today 2005, 109, 1–2. (4) Prins, R.; Egorova, M.; Röthlisberger, A.; Zhao, Y.; Sivasankar, N.; Kukula, P. Catal. Today 2006, 111, 84–93. (5) Long, F. X.; Gevert, B. S. J. Catal. 2004, 222, 1–5. (6) Alcaide, F.; Cabot, P. L.; Brillas, E. J. Power Sources 2006, 153, 47–60.
accounts for 48%.7 Reforming is a potential method for obtaining renewable hydrogen from crude glycerin. Reforming reactions are generally endothermic, and a reforming process may be characterized as steam reforming, catalytic partial oxidation, or autothermal reforming depending upon the source of heat and types of reactants.8 A general equation to describe glycerin reforming is shown in eq 1. C3H8O3 + xH2O + yO2 f aCO2 + bCO + cH2O + dH2 + eCH4 + ... (1) External heat may be provided to the process as in the case of steam reforming (eq 1: x > 0, y ) 0) or an oxygen source may be included to react with the feed stock and provide the required heat internally as in the case of catalytic partial oxidation (eq 1: x ) 0, y > 0) and autothermal reforming (eq 1: x > 0, y > 0). The amounts of H2O and O2 added as reactants are reported relative to the carbon input in the feed stock, i.e., molar ratios of steam to carbon (S) and O2 to carbon (O). In this case, O includes only O2 added in pure form or air and does not include oxygen present in steam or in the fuel. The equilibrium composition depends on these reactant ratios and the reaction temperature (T) and pressure. As indicated in eq 1, reforming products include hydrogen and carbon monoxide as (7) National Hydrogen Vision Meeting. A National Vision of America’s Transition to a Hydrogen Economy- To 2030 and Beyond. United States Department of Energy. www1.eere.energy.gov/hydrogenandfuelcells/pdfs/ vision_doc.pdf (accessed Feb 2002). (8) Joensen, F.; Rostrup-Nielsen, J. R. J. Power Sources, 2002, 105, 195–201.
10.1021/ef060379w CCC: $37.00 2007 American Chemical Society Published on Web 09/05/2007
3500 Energy & Fuels, Vol. 21, No. 6, 2007
well as carbon dioxide and methane.9 A catalyst is used to accelerate the reactions in the reforming process. Ni, Co, Ni/ Cu, and noble metal (Pd, Pt, Rh) based catalysts all favor hydrogen production with Ni being the most commonly used. Catalysts enhance the reforming reaction rates at the molecular level and thorough discussions of the topic are available in the literature.10–14 This paper presents results of experiments using glycerin as a reforming feed stock for hydrogen production. Crude glycerin obtained from biodiesel production may contain impurities such as NaCl, NaOH, methanol, and free fatty acids as a result of the biodiesel manufacturing process. Crude glycerin composition varies from batch to batch, and therefore, reagent-grade glycerin was initially used to remove the effect of contaminants from the experiments. Afterward, crude glycerin and reagent-grade glycerin doped with contaminants were tested and compared with reagent-grade glycerin results. Stoichiometric steam reforming provides an upper yield limit of 7 mol H/mol glycerin as shown in eq 2. This theoretical yield served as a benchmark for comparison with experimental results.
Douette et al.
Figure 1. Schematic of glycerin, nitrogen, and oxygen injection system. Diameters of the inner, middle, and outer tubes are 0.03, 6, and 12 mm, respectively.
C3H8O3 + 3H2O f 3CO2 + 7H2 ∆H298K ) 345 kJ/mol (2) Experimental Section A 20 L sample of crude glycerin byproduct was obtained from biodiesel produced from fat, oil, and grease recovered from food service activities (Pacific Biodiesel, Honolulu, HI). A subsample was subjected to proximate (volatile matter, ash, and fixed carbon), ultimate (C, H, O, N, S, ash), and heating value analyses. A sample of the ash was analyzed for the elements Si, Al, Ti, Fe, Ca, Mg, Na, K, and P. The bulk sample was also analyzed for selected metals including As, B, Cd, Pb, Mn, Hg, Mo, Se, and Zn. All analyses were performed by Hazen Research Inc. (Golden, CO). Reagent-grade glycerin obtained from Fisher Scientific (glycerin certified ACS, G33-1) was also used in the reforming tests. A laboratory scale reactor was used to study the effects of operating conditions on the yield of hydrogen obtained from reforming reagent-grade glycerin based on a 23 experimental design. The reforming reaction took place in a stainless steel pipe, 294 mm in length with an internal diameter of 25 mm. The reactor was placed inside a furnace to provide temperature control. Six pellets (19.6 g) of nickel-based catalyst (G-91 EW steam reforming catalyst, Süd-Chemie Inc., Louisville, KY) were placed in the reformer, 16 cm from the entrance. Previous experiments with liquified petroluem gas (LPG) indicated that this amount of catalyst was adequate for the carbon feed rate of ∼0.25 mol/min.10 Flows of a liquid phase glycerin and water mixture and gas phase oxygen and nitrogen were metered to the reformer. Air could have been used for this experiment, but operating experience showed it to be more practical to separate the nitrogen and oxygen flows. Glycerin, O2, and N2 were introduced to the reactor through three concentric tubes as shown in the Figure 1. Typically, a 38% glycerin and 62% water mixture was injected into the reformer through the 0.03 mm internal diameter center tube at a rate of 3 mL/min. The mixture (9) Song, X.; Guo, Z. Energy ConVers. Manage. 2006, 47, 560–569. (10) Wang, W.; Turn, S. Q.; Keffer, V.; Douette, A. Parametric Study of Authothermal Reforming of LPG. Prepr. Pap.–Am. Chem Soc., DiV. Fuel Chem. 2004, 481, 142. (11) Song, X.; Guo, Z. Energy ConVers. Manage. 2006, 47, 560–569. (12) Nurunnabi, M.; Mukainakano, Y.; Kado, S.; Li, B.; Kunimori, K.; Suzuki, K.; Fujimoto, K.; Tomishige, K. Appl. Catal., A: General 2006, 299, 145–156. (13) Hou, K.; Hughes, R. Chem. Eng. J. 2001, 82, 311–328. (14) Sehested, J. Catal. Today 2006, 111, 103–110.
Figure 2. Thermocouples and catalyst positions inside the reformer (distances in centimeters).
percentages were determined by first choosing the water to carbon molar ratio for the experimental conditions and then converting this ratio to a volumetric mixture for water and glycerin. Nitrogen (1.75 L/min) was introduced in the annulus between the glycerin injector and a 6 mm tube, and the oxygen was introduced into the annulus between the 6 mm tube and an outer 12 mm tube. The nitrogen acted as a shroud gas to prevent the oxygen from immediately reacting with the glycerin at the injector tip. Gas flows were metered using mass flow controllers (Model 5890e, Brooks Instruments, Hatfield, PA), and liquid flow was regulated using a precision pump (Model QG50, FMI Pump, Soyesset, NY). An in-line sintered stainless steel filter was located 125 mm downstream of the pipe to collect any particulate matter produced in the reformer and protect downstream analytical equipment. System pressure was monitored at the reactor inlet. As shown in Figure 2, three, type K thermocouples were located inside the reformer, two upstream of the catalyst and one downstream of the catalyst, to monitor reactor temperatures. The reactor operating conditions (pressure and temperature) were monitored and recorded every 3 s and any sudden changes in operating conditions were readily observed. The gas exiting the reformer passed through a condenser to remove water vapor and was then disposed. The reformate stream was sampled and analyzed using a GC (Shimadzu 14A) equipped with a thermal conductivity detector and packed column (Carboxen 1000, Supelco, Bellafonte, PA). The GC was equipped with an automatic sampling valve that injected a sample every 35 min. The sample loop was purged for 3 min prior to each injection to ensure that fresh reformate was sampled. A 23 factorial experimental design15 was used to study the effects of the oxygen to carbon ratio (O), steam to carbon ratio (15) Box, G. E. P.; Hunter, W.; Hunter, J. S. Statistics for Experimenters: An Introduction to Design, Data Analysis, and Model Building; John Wiley & Sons: New York, 1978.
Hydrogen Production from Glycerin Reforming
Energy & Fuels, Vol. 21, No. 6, 2007 3501
Table 1. Initial Reformer Conditions in Real and Coded Units variable values
Table 2. Reformate Gas Component Yields mol/mol glycerin (average ( standard deviation)
coded values
condition no.
O
S
T (°C)
X1
X2
X3
condition no.
H2
CO
CO2
CH4
1 2 3 4 5 6 7 8 9 10
0.4 0.7 0.4 0.7 0.4 0.7 0.4 0.7 0.55 0.55
2.0 2.0 2.7 2.7 2.0 2.0 2.7 2.7 2.4 2.4
770 770 770 770 850 850 850 850 810 810
-1 1 -1 1 -1 1 -1 1 0 0
-1 -1 1 1 -1 -1 1 1 0 0
-1 -1 -1 -1 1 1 1 1 0 0
1 2 3 4 5 6 7 8 9 10
3.2 ( 0.1 1.5 ( 0.5 3.1 ( 0.1 1.9 ( 0.2 3.5 ( 0.1 1.3 ( 0.1 3.5 ( 0.3 1.2 ( 0.1 2.1 ( 0.1 2.1 ( 0.05
1.0 ( 0.03 0.5 ( 0.2 0.8 ( 0.03 0.5 ( 0.3 1.2 ( 0.03 0.7 ( 0.07 1.0 ( 0.1 0.6 ( 0.05 0.8 ( 0.1 0.8 ( 0.04
2.0 ( 0.01 2.3 ( 0.3 2.0 ( 0.02 2.4 ( 0.2 1.9 ( 0.02 2.2 ( 0.02 2.0 ( 0.1 2.3 ( 0.05 1.9 ( 0.02 2.2 ( 0.04
0.03 ( 0.001 0.02 ( 0.006 0.027 ( 0.001 0.008 ( 0.001 0.04 ( 0.001 0.03 ( 0.002 0.04 ( 0.003 0.03 ( 0.003 0.04 ( 0.006 0.04 ( 0.008
(S), reformer temperature (T), and the interaction of the three variables on hydrogen yield and hydrogen concentration produced from reforming glycerin. Initial reforming conditions were chosen based on literature review and thermochemical equilibrium calculations. The initial condition for the three variables under study were O ) 0.55 (mol O2/mol C), S ) 2.35 (mol H2O/mol C) and T ) 810 °C. O was varied by (0.15 to 0.4 and 0.7, S was varied by (0.35 to 2.0 and 2.7, and the temperature was varied by (40 °C to 770 and 850 °C. Table 1 summarizes the values of the original test conditions including two center points (conditions 9 and 10). Coding was used to make the statistical analysis easier. In coded units, X1, X2, and X3 represent O, S, and T, respectively. A zero value represents the center point for the variable, and +1 and -1 represent the upper and lower values for the three variables under study. The performance of the experiment is defined primarily by studying the H2 yield (mol H2/mol glycerin) and also yields of CO, CO2, and CH4. On the basis of results from these initial experiments, response surface methodology was used to improve the process performance by varying S, O, and T. For each condition (1–10), the reformate gas was analyzed for the predominant gases, N2, H2, CO, CO2 and CH4, and the trace gases, C2H6 and C2H4. Gas yield was computed based on the measured nitrogen concentration in the reformate (GC analysis) and the known flow rate of nitrogen through the reactor. After optimizing the reforming process, a water gas shift reactor was added in series with the reformer to investigate its effect on hydrogen production. The experiments used a monolithic water gas shift (WGS) catalyst on a nanoparticle, ceriabased, mixed oxide support (NexTech Materials, Ltd., Lewis Center, OH). Temperature was the only variable changed in tests of the shift reactor performance, and the reformer was operated at the optimum conditions identified earlier. Tests were characterized by an initial transient period until steady experimental conditions were attained as indicated by stable pressure and temperature profiles within the reactor. After reaching steady state, the test was continued until six gas samples were obtained (∼3.5 h) or until it failed due to the reactor becoming plugged as indicated by increasing pressure. Each test condition indicated in Table 1 was conducted separately. After each test, the reactor was opened and cleaned and prepared with fresh catalyst for the ensuing test. Results and Discussion Reformer. Results of the initial 23 factorial design are presented in Table 2. H2 yield ranged from 1.2 to 3.5 mol H2/ mol glycerin. CO and CO2 yields ranged from 0.5 to 1.0 mol/ mol glycerin and 1.9 to 2.3 mol/mol glycerin, respectively. CH4 yields were less than 0.04 mol/mol glycerin for all test conditions. From the results, mathematical models describing the H2, CO, and CH4 yields (YH2, YCO, and YCH4, respectively)
as functions of the coded units for S, O, and T were derived; see eqs 3, 4, and 5. Experimental values of species yield plotted against the values predicted by the models are shown in Figure 3 and are in close agreement in all three cases. YH2 ) 2.3 - 0.93X1 + 0.034X2 - 0.02X3 + 0.06X1X2 - 0.18X1X3 - 0.047X2X3 - 0.08X1X2X3 (3) YCO ) 0.79 - 0.21X1 + 0.06X2 - 0.09X3 + 0.04X1X2 - 0.12X1X3 - 0.012X2X3 - 0.01X1X2X3 (4) YCH4 ) 0.27 - 0.0054X1 + 0.0011X2 - 0.006X3 + 0.0X1X2 - 0.0011X1X3 - 0.0019X2X3 - 0.001X1X2X3 (5) Equations 3 and 4 show that X1 (i.e., O in coded units), had the greatest effect on H2 and CO yields. H2 yield was also affected to a lesser extent by the interaction between O and T. The steam to carbon ratio had very little effect on the hydrogen yield. Methane yields were affected equally by O and T as shown in eq 5. Following surface response methodology, eq 3 was used to determine new values of O, S, and T (see Table 3) predicted to improve YH2 along a path of steepest ascent. Equation 3 shows that to improve YH2, O had to be decreased while almost keeping S and T at their center point values from the original test conditions. Table 4 shows the gas yields for reforming experiments conducted at conditions 11 and 12. By minimizing the oxygen input (a reduction of 52% from the center point), the hydrogen yield was improved from ∼2 to 4.5 mol H2/mol glycerin, i.e., 65% of the theoretical yield (7 mol H2/mol glycerin) in eq 2. A 1.4 mol portion of CO was also produced, which could theoretically produce another 1.4 mol of H2 by the water gas shift reaction as shown in eq 6. CO + H2O f CO2 + H2
∆H298K ) –41 kJ/mol
(6)
The amount of carbon monoxide increased in comparison to conditions 1–10, showing that the reduced O2 input produced more CO instead of CO2. Figure 4 shows the reformate hydrogen and carbon monoxide concentrations for the 12 test conditions. The experimental results were compared against the equilibrium composition predicted using FactSage, a thermochemical software and database package. At the values of O, S, and T used for condition 12, thermochemical equilibrium predicted 5.4 mol H2/mol glycerin. The experimental value of 4.5 mol H2/mol glycerin thus achieved 83% of equilibrium. Water Gas Shift. The water gas shift optimal temperature was reached following the same surface response approach used to improve the reformer operation. Figure 5 shows the H2 yields obtained operating the reformer and the water gas shift in series. The hydrogen yield per mole of glycerin increased from 4.5 mol H2/mol glycerin exiting the reformer to 5.3 mol H2/mol glycerin when the WGS was used, an improvement of 18%. Also, the amount of CO2 rose from 1.4 to 2.2 mol H2/mol glycerin, and the CO decreased from 1.4 to 0.5 mol H2/mol
3502 Energy & Fuels, Vol. 21, No. 6, 2007
Douette et al.
Figure 3. H2, CO, and CO2 yield: actual vs predicted value. Table 3. Experimental Conditions Determined from Response Surface Analysis Variables condition no.
O
S
T (°C)
11 12
0.3 0
2.3 2.2
807 804
Table 4. H2, CO, and CO2 Yields from Conditions along Path of Steepest Ascent mole/mole glycerin condition no.
H2
CO
CO2
11 12
3.8 ( 0.07 4.5 ( 0.08
1.0 ( 0.03 1.4 ( 0.04
1.9 ( 0.02 1.4 ( 0.02
glycerin. These results are in keeping with the water gas shift reaction (eq 6), since the CO2 and H2 production each increased by 0.8 mol and the CO production decreased by 0.9, showing that approximately all the CO is converted to CO2. If the water gas shift was the only reaction taking place in the shift reactor, an additional 0.1 mol of CO2 should have been produced to complete the carbon balance. The gas analysis showed that 0.065 mol of C2H6 was produced per mole glycerin, compared to 0.01 mol/mol glycerin at the reformer exit for condition 12. This could account for the 0.1 difference between the CO and CO2 yields. The carbon balance computed for the final condition was 97.6% indicating that the measurement of system inputs and outputs are in close agreement. The addition of the shift reactor to the reformer improved hydrogen yield from 64% to 75% of the maximum theoretical yield. Catalyst Deactivation. Catalyst deactivation occurred over time mainly due to coking and resulted in a lower hydrogen
yield.10,16–20 An indicator of catalyst deactivation is an increase in methane production, and to a lesser extent, ethane. A decrease in reformer performance was noticeable after a few hours of testing. For condition 12 (i.e., best reforming condition), the hydrogen yield decreased from 4.8 to 4.5 mol H2/mol glycerin during the first three hours, a reduction of at 2.1% per h. Figure 6 shows the hydrogen yield calculated from successive GC samples at 35 min intervals. During the same time period, the methane production increased from 0.1 to 0.16 mol CH4/mol glycerin, or 20% per h. The effects of S and O on methane production (and catalyst deactivation) were studied using the data from the original design (i.e., conditions 1–8). Similar levels of methane production were found for both low and high S values. Comparing the methane production for the low and high O values showed a 30% lower methane production for the higher oxygen input. For O ) 0.4, YCH4 ) 0.033 mol CH4/mol glycerin, and for O ) 0.7, YCH4 ) 0.023. At condition 12, where O ) 0 (no external O2 was supplied), methane production was 0.1 mol/ mol glycerin. The lower methane production at higher oxygen to carbon ratios is explained by the fact that supplemental oxygen helps prevent carbon deposition on the catalyst surface thereby maintaining catalyst activity.10,17 Carbon deposits were readily (16) Cheekatamarla, P. K.; Lane, A. M. J. Power Sources, 2005, 152, 256–263. (17) Hardiman, K. M.; Cooper, C. G.; Adesina, A. A.; Lange, R. Postmortem characterization of coke-induced deactivated alumina-supported Co–Ni catalysts. Chem. Eng. Sci. 2006, 61, 2565–2573. (18) Rakass, S.; Oudghiri-Hassani, H.; Rowntree, P.; Abatzoglou, N. Steam reforming of methane over unsupported nickel catalysts. J. Power Sources 2006, 158, 485–496. (19) Akande, A J.; Idem, R. O.; Dalai, A. K. Appl. Catal., A: General 2005, 287, 159–175. (20) Melo, F.; Naphtha, N. M. Catal. Today 2005, 107–108, 458–466.
Hydrogen Production from Glycerin Reforming
Energy & Fuels, Vol. 21, No. 6, 2007 3503
Figure 4. H2 and CO concentrations (vol %, dry basis) in reformate gas for conditions 1–12 (see Tables 1 and 4 for experimental conditions associated with the numbers in the legend).
Figure 5. Hydrogen yield as a function of water gas shift temperature with the reformer operated at condition 12.
apparent on the catalyst pieces closest to the reformer inlet when they were recovered after a test was completed. Catalyst pieces farthest downstream kept their original, grayish color. The amount of carbon recovered from the reformer and the filter after each test was small (0.36–1.0 g), accounting for less than 1% of the total carbon contained in the glycerin feed. A secondary impact of carbon formation was an increase in reformer operating pressure over the test duration. A higher S/C could be expected to reduce coking and slow the catalyst deactivation. From the amount of carbon collected in the reformer and the pressure recorded in the reformer, it was not possible to determine a clear effect of S/C on the carbon formation for the duration of these tests. The label of the reagentgrade glycerin used in the reforming tests indicated that sulfur (as sulfate) impurities could also be present at levels of
