Numerical Study on Catalytic Hydrodeoxygenation of Pyrolytic Bio-Oil Model Compound, Guaiacol, in Fluidized Bed Reactor

The bio-oil obtained by thermochemical conversion of lignocellulosic biomass consist of large fractions of oxygenated compounds which deteriorate its quality leading to low calorific value, high viscosity, high density, high moisture content, etc. Therefore, the bio-oil should be deoxygenated using hydrogen in the presence of appropriate catalyst to improve its properties. Adequate literature on pyrolysis of biomass within the framework of computational fluid dynamics is available but only a couple of papers available on hydrodeoxygenation of bio-oil obtained by pyrolysis. Thus, in this study, guaiacol has been selected as a representative model compound of phenolic fraction of bio-oil for upgrading it by catalytic hydrodeoxygenation. The reaction process has been implemented in a fluidised bed reactor in the presence of palladium catalyst, Pd/Al 2 O 3 using computational fluid dynamics (CFD) based solver, ANSYS Fluent 14.5. The range of conditions considered herein are: weight-hourly space velocity (WHSV) = 1, 3 and 5 h -1 ; superficial H 2 -gas velocity, u = 0.075, 0.15 and 0.25 m/s; catalyst load = 0.06 kg and temperature, T = 548 K, 573 K, and 598 K. The solver has been thoroughly validated in terms of grid dependence study, time step size dependence study validating hydrodynamics and HDO results wherever possible with existing literature results. The HDO of guaiacol produces phenol as the most abundant compound along with significant amount of cyclopentanone and methanol. The formation of cyclopentanone from HDO of guaiacol is favourable at high temperature whereas low temperature conditions favour formation of methanol and phenol.


Introduction
Broadly speaking many sectors in the world today depend on energy sector as a major impetus upon which majority of their activities coordinate. The transport sector being the leading one and is estimated to consume approximately one-fifth of the total produced global energy [1]. This rate however, is likely to accelerate even further due to population growth, improved standards of living and the drive for economic growth by many developing economies. The non-renewable conventional fossil fuels (oil, gas and coal) are the most dependable global energy resources. Oil is the leading one and accounts for one-third of the global energy consumption [2]. Although it is economically viable to extract energy from fossil fuels compared to renewable energy resources, fossil fuels are detrimental to the environment through greenhouse gases emission especially CO2.
Fossil fuels are non-renewable thus high dependence on them may result into depletion albeit new conventional reserve discoveries and improved technologies currently available to extract fossil fuels even from tight non-conventional reserves. Because of above-mentioned factors and among others there is relentless shift in attention towards cleaner and lower carbon alternative or renewable energy resources that are environment friendly. Statistical review carried out by British Petroleum, BP in 2017 [2], shows that renewable energy including biofuels despite having the least share of 4% was the fastest growing energy accounting almost a third of increase in primary energy. Similarly, according to BP energy outlook 2030 [3] it is estimated that renewable energy will contribute 17% increase in global energy supply by the year 2030. Therefore, the future for global energy is renewable resources with bio-oil obtained from biomass-waste being the most promising alternative fuel to petroleum fuels.
There are several approaches available to convert biomass into bio-energy out of which thermochemical conversion methods have wide acceptance [4,5]. The processes such as gasification, combustion, pyrolysis and liquefaction fall under thermochemical conversion processes. Pyrolysis method has been extensively researched in order to produce bio-oil from lignocellulosic waste biomass. However, the bio-oil obtained by pyrolysis suffer from severe drawbacks such as low calorific value, high density, high viscosity, high moisture content, low pH, etc. This is mainly due to presence of large fractions of oxygenated compounds. Thus by removing these oxygen-fraction, the quality of bio-oil improves in terms of calorific value, density, viscosity, moisture content, etc. much close to the quality commercial gasoline, diesel and kerosene range fractions. Catalytic hydrodeoxygenation is one of the most appropriate approach to remove oxygenated fractions from the compounds of bio-oil using hydrogen in the presence of an appropriate catalyst. Furthermore, with advancement in computation, CFD has extensively been used in modelling, design, predicting, optimization and improving a number of chemical processes while reducing costs associated with experiments and other unclassified expenditures. A legion of researchers has used CFD successfully to study hydrodynamics as well as chemical conversions of processes involving biomass pyrolysis, gasification, coal combustion and fluid catalytic cracking. In the recent past, a couple of studies have also used CFD for modelling and simulations of hydrodeoxygenation of bio-oil model compounds using lumped kinetic parameters so that to gain insight into the processes and to provide scale-up solutions [6,7]. Therefore, in this study, a model compound of phenolic fraction of bio-oil, guaiacol, has been considered to study its hydrodeoxygenation behaviour in a fluidized bed reactor. Wide range of conditions have been chosen namely: weight-hourly space velocity (WHSV) = 1, 3 and 5 h -1 ; superficial H2-gas velocity, u = 0.075, 0.15 and 0.25 m/s; catalyst load = 0.06 kg and temperature, T = 548 K, 573K and 598 K.

Literature Review
The literature on CFD modelling and simulation of hydrodeoxygenation of bio-oil is limited to recent studies of Subramanyam et al. [6] and Gollakota et al. [7]; hence literature pertaining to CFD applied to pyrolysis, gasification, combustion, fluid catalytic cracking, etc. are presented herein. In an attempt to explore and improve computation research carried out in modelling thermochemical conversion of biomass ,C.Di Blasi [8] reviewed numerous previous reports in state of the art modelling of biomass pyrolysis. The author described chemical kinetics, activation energy models as well as different approaches used in particle transportation models. Nonetheless, the author cited need for more experimental data to validate models of different reactors.
Bruchmu¨ller et al. [9] presented a DEM/CFD 3-D model using an Eulerian/ Lagrangian approach to study thermochemical degradation process of biomass inside an experimental 100 g/h lab scale bubbling fluidised bed reactor .Thorough analysis of the 3-D simulations indicated that fast pyrolysis of biomass to yield bio-oil is vast influenced by local flow of the biomass influenced by the superficial fluidisation velocity rather than the particle such properties as moisture content.
Xue et al. [10] developed an Euler-Euler CFD computation model coupling pyrolysis of biomass particles with multi-fluid hydrodynamics model for gas-particle flow. The author to study biomass pyrolysis process in a fluidized bed reactor implemented this model. Experimental validation of the model with data demonstrated accurate qualitative and quantitative capability to describe complex conversion of biomass and transport processes. Similarly, Ranganathan and Gu [11] carried out CFD modelling of biomass fast pyrolysis using the same reactor as Xue et al [10]. The author first studied hydrodynamics of the reactor and later investigated different kinetic schemes for fast pyrolysis of biomass of different particle density and size at various gas velocity.
Papadikis et al. [12] investigated and compared two cases involving 2-D and 3-D geometry approaches for modelling momentum transport in a 150 g/h lab scale fluidised bed reactor. The author obtained different results based on the different geometry approaches.
However, due to computation limitations exhibited by 3-D geometry the author generally recommended 2-D geometry for research purposes. Nevertheless, Papadikis et al. [13] continued the same research modelling fluid-particle interaction in the same reactor further incorporating reaction kinetics of biomass with the properties of the moving discrete particle of the biomass. The author investigated the effect of heat and mass transport of the fluidising gas and reported that the model predicted residence time of the vapours and the particles. Mellin et al. [14] carried out CFD modelling of biomass pyrolysis incorporating a complex reaction scheme including formation of components such as levoglucason. Results predicted that pyrolysis products reflected the experimental yield satisfactorily. Similarly, the author noted complete conversion of hemicellulose than cellulose and lignin. Lee et al. [15] analysed the effect of reaction temperature on reaction rate and final product yield using two simulation models viz lumped model and hybrid model. It was noted the yield of tar from the hybrid model prediction was consistent with experimental results than the lumped model. However, for char yield both model results obtained from simulation were close to those from the experiment. Notwithstanding the application of CFD in the study of the above processes it also has been applied in hydrodynamic analysis and study of fluidised catalytic cracking (FCC) reactions.
Soundararajan et al. [26] modelled methanol to olefin process in a fluidised bed reactor at 450 o C and atmospheric pressure. Modelling studies by the author showed that selectivity towards ethylene increased significantly with increase in coke deposit on the catalyst. However, higher amount of coke deposition on the catalyst beyond 5 wt% inhibits the increase in ethylene yield.
Similarly, Zhuang et al. [27] simulated coke deposition and distribution during methanol to olefin reaction as a function of feed temperature, feed composition and space velocity over SAPO-34 catalyst in a fixed bed reactor. It was noted higher feed temperature could promote methanol conversion and accelerate catalyst deactivation. Chang et al. [28] investigated on hydrodynamics and kinetic reactions in a fluidised bed methanol to olefin reactor. Results from simulation indicated that the rate of conversion of methanol and product yields are more sensitive to reaction and pressure than initial methanol content in the feedstock.
Behjat et al. [29] developed a 3-D CFD model of a reactor to study cracking reactions, hydrodynamics ,heat and mass transfer of a three phase (gas-liquid-solid) six-lump reaction scheme. The author further investigated the evaporation tendency of the feed droplets into a gassolid flow. It was found evaporation of gas oil droplets and cracking reactions have profound effect on gas-solid flow and temperature. Similarly, Chang et al. [30] used CFD to study hydrodynamics and cracking reactions in a heavy oil riser. The author reported that product yields are more sensitive to injection angle into the riser than droplet size of the feedstock and reaction temperature. Nonetheless, Yang et al. [31] developed a three-dimensional model to investigate hydrodynamics, heat transfer and cracking reaction to maximise propylene production implementing an 11-lump kinetic model. The author analysed effects of injecting angle of the nozzle on flow and thermal patterns at the feed injection zone. It was recommended wider angles of injection induce high radial velocities, better heat transfer and desirable reaction performance.
Recently, Pelissari et al. [32] also simulated the hydrodynamics, heat transfer and heterogeneous catalytic cracking reactions within a riser using a 12-lump kinetic model. The author proposed a catalyst deactivation model dependent on the weight percentage of coke amount on the catalyst.
Analysis of results indicate treatment for coke has significant role in simulation with catalyst deactivation as a function of coke amount on catalyst.
Finally, based on the aforementioned literature survey, it is clear that much emphasis has been focused on CFD applied to biomass pyrolysis, coal combustion, gasification and fluid catalytic cracking. However, CFD applied to hydrodeoxygenation of bio-oil has received negligible attention except those of Subramanyam et al. [6] and Gollakota et al. [7]. Thus, HDO of bio-oil has been carried out using a representative bio-oil model compound of phenolic group, guaiacol. The reaction has been implemented in a fluidised bed reactor using palladium catalyst objectively to study the effect of temperature, weight hourly space velocity and superficial velocity of hydrogen gas.

Problem Statement, HDO Kinetics and Mathematical Formulation
The schematic illustration of a fluidized bed reactor used for the hydrodeoxygenation of guaiacol is shown in Figure 1 which is of height (H) 1.0m and diameter (d) 0.28m. The reactor is packed up to an initial packing height (H0) of 0.4m corresponding to solids volume fraction of 0.6. Authors have chosen this size of reactor for the validation purpose (without hydrodeoxygenation reaction) with the study of Taghipour et al. [33] who have reported experimental and computational results on fluidization behaviour of glass beads as solid particles of size 275µm and density 2500kg/m 3 .
The validation part is shown in a subsequent section. For the case of HDO of guaiacol, the same reactor size having same initial packing bed height, size of packing material and solid bed volume fraction as in the case of Taghipour et al. [33] has been used. However, the packing material type is now mixed Pd/Al2O3 catalysts and glass beads while the fluidized medium is hydrogen gas along with the bio-oil model compound, guaiacol. Alternatively, it can be said that the hydrogen gas and guaiacol are introduced in to the reactor at different flow rates where the guaiacol flow rate is computed depending on calculated based on the required value of the weight hourly space velocity.
The temperature inside the reactor is maintained constant for a given run and three different temperatures are considered for which kinetics of guaiacol HDO are available due to experimental work of Gao et al. [34]. For specified hydrogen gas superficial velocity, reactor temperature and bio-oil mass flowrate per unit mass of catalyst (i.e., WHSV) and corresponding kinetics at that temperature, different degree of HDO of guaiacol take place inside the reactor producing phenol, catechol, cyclopentanone, methanol, water and gases which occupy the freeboard of the reactor in addition to interstitial spaces between the catalyst particles. The performance of HDO would strongly be affected with the distribution of solids in the reactor while fluidization is occurring in addition to momentum and heat transfer amongst three different phases. Thus, species involved in this study are grouped in to three phase mixtures, i.e., bio-oil phase mixture, gas-phase mixture and solid-phase mixture. The thermo-physical properties of these phases can be in Table 1. The bio-oil mass flow rate into the reactor is determined based on the value of weight hour space velocity. Flow conditions of both bio-oil model compound and hydrogen gas superficial velocity used in this study are summarized in Table 2. The catalyst's volume fraction inside the reactor is very small (only 0.0002) compared to the volume of the reactor. Thus, in order to make the bed height of up to 0.4m as well as increase the solids bed density in the present HDO reactor, glass beads of the same size and density same as used in the study of Taghipour et al. [33] was added to the catalyst bed. The solids volume fraction αs inside the reactor was determined by expression

Effect of grid size, time step and validation
Literature results by Taghipour et al. [33] was used to validate the hydrodynamics of a non-reacting gas-solid fluidized bed system by maintaining identical conditions, i.e., using air as fluidizing medium. The other identical conditions include that the reactor was packed with solids (glass beads) of density 2500 kg/m 3 and mean diameter of 275µm to an initial static bed height H0 = 0.4m corresponding to solids volume fraction 0.6. In other words, the study of Taghipour et al. [33] replicated here for the purpose of validation. Then in the later course (as discussed in subsequent subsections of results), the HDO reaction of guaiacol is incorporated using hydrogen gas and bio-   Unlike the other three compounds, formation of catechol in the mixture depends on WHSV and varies differently at various temperatures. It can be seen at T=548 K and T=573 K, catechol accordingly exhibits high and low mass fraction at WHSV=1 h -1 (guaiacol flow rate). However, as WHSV increases mass fraction of catechol at T=573 K increases steadily while at T=548 K decreases with minimum fraction of the compound depicted at WHSV=3 h -1 . A consistent similar observation at T=548 K is also seen at T=598 K. Although catechol fraction in the mixture increases with guaiacol flow rate, generally the compound has uniformly very low or none fraction compared to other HDO products of guaiacol. Comparing results obtained at another H2-gas superficial velocity, Figure 5 shows that product distribution of guaiacol HDO for instance at u=0.25 m/s is somewhat analogous to u=0.075 m/s notwithstanding disparity in results of mass fraction of catechol. In fact at u=0.25 m/s catechol formation in the mixture occurs only at moderate temperature T=573 K.

HDO of Guaiacol
By varying H2-gas velocity at T= 548 K as indicated in Figure 6, both methanol and phenol exhibit a consistent pattern in variation of mass fraction opposite to cyclopentanone. It is typically observed that as H2-gas velocity progressively increases from 0.075 to 0.25 m/s methanol fraction at WHSV (= 1 h -1 ) decreases and at WHSV=5 h -1 increases with an equal percentage of 0.12 %.
Similar observation is depicted with phenol however, with a corresponding factor of 0.04 %.
Generally high H2-gas superficial velocity (u=0.25 m/s) favours formation of methanol and phenol at low WHSV and favours cyclopentanone formation at high WHSV (=5 h -1 ). Catechol shows corresponding decrease and increase in mass fraction but at WHSV= 1 and WHSV=3h -1 respectively as H2-gas velocity increases.
At T=598K, increasing H2-gas superficial velocity shows that mass fraction of respectively. There is very little or no catechol formed. Hydrodeoxygenation of guaiacol produces a range of products. From the reaction progress it can be said guaiacol HDO proceeds via direct demethoxylation to phenol with subsequent formation of methanol and direct hydrogenation forming cyclopentanone as proposed by Gao et al. [34].

Conclusions
The hydrodeoxygenation of guaiacol has been carried out in a fluidized bed reactor within the framework of computational fluid dynamics over wide ranges of hydrogen superficial velocity,

Conflicts of interest/Competing interests (include appropriate disclosures):
• Herewith authors declare that there is no conflict of interests of their work reported in this manuscript

Availability of data and material (data transparency):
• Data and equations related to this work have been included in the supplementary information of the manuscript

Code availability:
• This work has been carried out using ANSYS Fluent software and the plotting of figures have been done using MS Office Excel software. The manuscript has been prepared using

MS Office Word software
Authors' contributions: • Mr. Ogene Fortunate has done the simulations using the ANSYS Fluent and prepared a first draft • Prof. Nanda Kishore has designed the problem, received funding for the project, supervised the project, edited and corrected the manuscript and checked similarity index using Turnitin.com which reported only 12% similarity.   Table 3 Activation energy (Ea,i) and pre-exponential factors (Ai) for HDO of guaiacol at different operating temperatures [34].