Energy & Fuels 1998, 12, 19-24
19
Production of Hydrogen from Biomass by Catalytic Steam Reforming of Fast Pyrolysis Oils Dingneng Wang, Stefan Czernik, and Esteban Chornet* National Renewable Energy Laboratory (NREL), Golden, Colorado 80401 Received July 2, 1997. Revised Manuscript Received September 23, 1997X
Hydrogen is of great interest as the cleanest fuel for power generation using fuel cells and for transportation. Biomass can be thermochemically converted to hydrogen via two distinct strategies: (1) gasification followed by shift conversion, and (2) fast pyrolysis of biomass followed by catalytic steam reforming and shift conversion of specific fractions. This paper presents the latter route. The process begins with fast pyrolysis of biomass to produce bio-oil, which (as a whole or its selected fractions) can be converted to hydrogen via catalytic steam reforming followed by a shift conversion step. Such a process has been demonstrated at the bench scale using model compounds and the aqueous fraction of poplar oil with commercial nickel-based steam-reforming catalysts. Hydrogen yields as high as 85% of the stoichiometric value have been obtained. Initial catalyst activity can be maintained through periodic regeneration via steam or carbon dioxide (CO2) gasification of the carbonaceous deposits.
Introduction Renewable lignocellulosic biomass has been considered as a potential feedstock for gasification to produce syngas (a mixture of hydrogen and carbon monoxide) for the last few decades. However, the economics of current syngas production processes favor the use of hydrocarbons (natural gas, C2-C5, and naphtha) and inexpensive coal. An alternative approach to the production of hydrogen from biomass is fast pyrolysis of biomass to generate a liquid product (also known as biooil) and catalytic steam reforming of the whole oil or its fractions. This approach has the potential to be cost competitive with the current commercial processes for hydrogen production. This work is designed to prove the feasibility of a process strategy that would use biomass, of either agricultural or forest origin, as a feedstock for producing bio-oil in small and medium size (99%). The H2 yields for all catalysts and model compounds were high, averaging 90% ((5%) of the stoichiometric value (Figure 3). Within our experimental error limit, there was no clear indication of one catalyst being better than the others. Next, we screened other reaction parameters and attempted to establish the best conditions to operate steam reformers at larger scales.6 Among the most important parameters for steam reforming are catalyst bed temperature (T), molar steam-to-carbon ratio (S/ C), methane-equivalent gas hourly space velocity (GC1HSV), and residence time (t, calculated from the void volume of the catalyst bed divided by the total flow rate of gases at the inlet of the reactor; void fraction ) 0.4). Temperature was found to have the most profound effect on steam-reforming reactions. Within experimental error limits, varying residence time from 0.04 to 0.15 s and increasing S/C from 4.5 to 7.5 showed no statistically significant effects on the yield of hydrogen under the conditions of 600 °C and GC1HSV ) 1680 h-1; however, these affected the concentration of CH4 in the product gas.
S/Ca
GC1HSVb
% C-to-gas conversion
% st. yield of H2 (+WGS) f
time on stream (h)
19 30 35 24 12 7 5 10 23 15 10
1110c 1000d 860c 850d 860d 860d 1010c 1230d 870e 850e 760e
97 102 99 100 97 98 92 102 92 89 86
103 (108) 103 (108) 86 (88) 89 (92) 82 (88) 79 (90) 75 (85) 78 (86) 73 (76) 71 (75) 71 (76)
2 4 1 1 1 1 6 6 1 1 4
a Molar ratio of steam to carbon. b Gas hourly space velocity on C1 basis (h-1) with ICI 46-1/4 catalysts. c Fresh catalyst used. d Used the same catalyst as the previous run above it. e UCI G-91 catalyst. f Assuming all CO being converted to H2 in a downstream WGS unit.
From these rapid screening studies of various classes of model oxygenate compounds, we concluded that oxygenate steam reforming faces a serious competition from gas-phase thermal decomposition prior to entering the catalyst bed and acid-catalyzed reactions at acidic sites in the catalyst support. These competing thermal cracking reactions may result in the formation of carbonaceous materials (coke) that can plug the reactor and deactivate the catalyst. If char formation prior to reaching the catalyst bed and carbonaceous deposits on the catalyst can be eliminated, or at least controlled, a complete conversion of both the oxygenate feed and its decomposition products to hydrogen can be achieved with commercial Ni-based catalysts under reasonable operating conditions. Therefore, it was decided that a special emphasis should be placed during bench-scale experiments on how to feed bio-oil or its fractions into the reactor. We conducted tests at the bench-scale level to obtain the global and elemental mass balances, yield of hydrogen, and the carbon-to-gas conversion. We quantified the distribution of gas products under conditions of complete conversion of the pyrolysis oil feedstock and studied the catalyst lifetime and the efficiency of its regeneration. We used model compounds (methanol, acetic acid, syringol, and m-cresol, both separately and in mixtures) and bio-oil (whole oil and its aqueous fraction). Representative results are listed in Tables 1 and 2. Time profiles of the output gas composition are shown for the three-component mixture in Figure 4 and for the poplar oil aqueous fraction in Figure 5. The following discussions are focused on the reforming of a three-component mixture and a bio-oil aqueous fraction.
Production of Hydrogen from Biomass
Figure 4. Composition of gaseous products during the steam reforming of a three-component mixture (67% acetic acid, 16% m-cresol, and 16% syringol) using the ICI 46 series catalysts (H2 (9), CO2 (O), CO (+), CH4 (4)). Conditions were 700 °C, S/C ) 5, GC1HSV ) 1050 h-1, and t ) 0.09 s.
Figure 5. Composition of gaseous products during the steam reforming of poplar oil aqueous fraction using the ICI 46 series catalysts (H2 (9), CO2 (O), CO (+), CH4 (4)). Conditions were 700 °C, S/C ) 30, GC1HSV ) 1000 h-1, and t ) 0.02 s.
The much lower CO concentration in the bio-oil aqueous fraction run is a result of the excess amount of steam used (S/C ) 30). The three-component mixture contained 67% acetic acid, 16% m-cresol, and 16% syringol. Its composition was close to the proportions of the carbohydrate fraction and the lignin fraction in bio-oil. We observed carbonaceous deposits on the top portion of the UCI G-90C catalyst bed when taking the reactor apart after the run. The overall mass balance (carbon, hydrogen, and oxygen) was 99% and the carbon conversion to gas was 96% (Table 1). The other catalyst system tested for steam reforming of the three-component mixture was the 46 series from ICI Katalco (46-1/46-4). This dual catalyst bed is used in commercial naphtha-reforming plants to reduce coke formation and extend catalyst lifetime. It showed an excellent and steady performance without any carbonaceous deposit on the catalyst. The gas composition (Figure 5) remained constant throughout the whole run. The overall mass balance (including carbon, hydrogen, and oxygen) was 104%, and for carbon it was 105%, indicating a small systematic error in our measurements. We obtained an excellent hydrogen yield of 86%, and the total hydrogen potential was as high as 98% with a second WGS reactor. These results
Energy & Fuels, Vol. 12, No. 1, 1998 23
Figure 6. Hydrogen yield profile before and after the regeneration cycle (by steam for 12 h) in a steam reforming experiment of poplar oil aqueous fraction using the ICI 46 series catalysts.
confirm that both the UCI G-90C and especially the ICI 46-series catalysts can efficiently convert oxygenates to hydrogen. Steam reforming of bio-oil or its fractions was found to be more difficult than that of model compounds. The main problem was feeding the oil to the reactor. Biooil cannot be totally vaporized; significant amounts of residual solids are often formed that block the feeding line and the reactor. Thus, the simple injection system used for model compounds had to be modified so that bio-oil and its fractions could be sprayed into the catalytic reactor without prior char formation. The aqueous fraction of NREL-made poplar oil was successfully fed to the reactor using a triple-nozzle spraying system with minimal accumulation of char in the reactor inlet. A large excess of superheated (850 °C) steam (S/C ) 20-30) was used, together with a high flow rate of nitrogen, to allow for proper oil dispersion and heat transfer required to maintain a sufficiently high temperature (>500 °C) at the reactor entrance. A portion of water and other volatiles in the sprayed droplets evaporated during mixing with the superheated steam and the remaining nonvolatiles contacted the catalyst surface directly. The ICI 46-series catalysts performed satisfactorily with no coke formation. We observed very stable gas composition and a constant production rate throughout the 4 h experiment (Figure 5). The carbon conversion of the aqueous fraction to gas products was almost quantitative in both runs that used the same catalyst bed (Table 2). We observed similar levels of mass balance closures as in the experiments using model compounds: global 99%, carbon 105%, and hydrogen 97%. The methane concentration (with N2 excluded) increased from 0.56% in the first run (2 h, t ) 0.03 s) to 2.2% in the second run (4 h, t ) 0.02 s). Both values were much higher than that (0.01%) obtained from the three-component model compound mixture (17 h, t ) 0.09 s). This was likely caused by the shorter residence time forced by the large steam and nitrogen flow rate used in the experiment. Lowering steam-to-carbon ratio caused an increase in CO yield and a decrease in H2 and CO2 yields, resulting from the WGS reaction. We also observed some thermal
24 Energy & Fuels, Vol. 12, No. 1, 1998
Wang et al.
regeneration by steam are also shown in Figures 6 and 7, indicating the recovery of H2 yield after a 12 h treatment at 800 °C and ambient pressure. Summary and Conclusion
Figure 7. Hydrogen yield profile before and after the regeneration cycle (by steam for 12 h) in a steam reforming experiment of poplar oil aqueous fraction using the UCI G-91 catalyst.
decomposition of the bio-oil fraction, which resulted in the accumulation of carbonaceous deposits in the region between the injection nozzle and the catalyst bed. There was a progressive formation of carbonaceous deposits in the catalyst bed. As a result, the region above the catalyst bed was totally blocked and the hydrogen yield gradually decreased (Figures 6 and 7). Therefore, it is necessary to regenerate the catalyst in order to achieve high performance over a long period of time. This can be effectively achieved by gasification with either CO2 (a coproduct of the process) or steam. Experiments using both model compounds (e.g., acetic acid) and the poplar oil aqueous fraction validated these methods of regeneration. The H2 yield profiles before and after
1. Biomass can be converted to bio-oil using fast pyrolysis technology. The yield of bio-oil for a fluidized bed process is on the order of 75 wt % (on dry biomass basis). 2. Bio-oil or its aqueous fraction can be efficiently reformed to generate hydrogen by a thermocatalytic process using commercial, nickel-based catalysts. The hydrogen yield is as high as 85% of the stoichiometric value. 3. Catalysts can be easily regenerated by steam or CO2 gasification of carbonaceous deposits. 4. The proposed strategy can be applied to any lignocellulosic biomass, either from agriculture or from forestry operations. Residues as well as dedicated energy crops are equally well suited as feedstocks to the process. The concept of using biomass via a regionalized system of production of bio-oils is entirely compatible with our strategy. Acknowledgment. The authors are indebted to the U.S. Department of Energy Hydrogen Program, under contract DE AC 36-83CH10093, for financial support. Technical discussions with Cathy Gre´goire-Padro´ and Margaret Mann and scientific contributions from Daniel Montane´ are acknowledged. EF970102J
