(738c) Networks with Parallel and Sequential Reactions for Determining the Pyrolysis Kinetics of Biomass Feedstocks
The design of optimal reactors for biomass pyrolysis requires a detailed kinetic model for every feedstock. To address this challenge, wewill first present a systematic procedure that can be used to develop lumped models for biomass pyrolysis. Earlier studies modeled biomass pyrolysis with two or three parallel reactions and used the distributed activation energy model (DAEM) to determine the kinetics of pyrolysis. Since each pseudo-component is actually a blend of decomposing species, the activation energies of the parallel decomposition reactions were assumed to be probability density functions with normal distributions. The parameters of these density functions (means, standard deviations, pre-exponential factors) were typically estimated by fitting data from thermogravimetric experiments with constant heating rates.
There is ample experimental evidence, however, that more complicated reaction networks must be used to describe biomass pyrolysis. In a classical paper, Keiluweit and coworkers (Environ. Sci. Technol. 2010, 44, 1247) summarized the physical and chemical transitions that biochars undergo as charring temperature increases from 100 to 700 °C. Starting from amorphous chars at low pyrolysis temperatures, a transition to composite chars with poorly ordered graphene stacks embedded in amorphous phases takes place as processing temperatures increase. Finally, turbostratic chars dominated by disordered graphitic crystallites are observed at higher temperatures.
To better capture this transition continuum, we developed a new framework that describes pyrolysis with a network of parallel and sequential reactions. While many earlier studies based their analysis on thermogravimetric experiments with constant heating ramps, our approach begins with step isothermal experiments aimed at identifying the potential pseudo-components of each biomass feedstock. Once the pseudo-components A1, A2, â¦, AN are identified, several reaction networks of the following form are built to model the pyrolysis of each feedstock.
Aj â Bj + Vj (1)
where Bj, j= 1,2,â¦,N are solid intermediates and Vj, j= 1,2,â¦,N are volatile species.
Some (or even all) of the intermediate products will further pyrolyze to give highly crystalline char C and moÂre volatiles according to:
Bk â C + Vk (2)
for some 1 â¤ k â¤ N.
Several thermogravimetric experiments with different heating ramps are then carried out for each biomass feedstock. The transient mass balances describing decomposition of all pseudo-components and intermediates according to reactions (1) and (2) yield a system of ordinary differential equations that are integrated numerically together with the temperature program used for the thermogravimetric experiments. Finally, nonlinear least squares techniques are used to fit model predictions to experimental data and estimate the kinetic parameters (activation energies, pre-exponential factors, orders of reaction etc.) for each pyrolysis reaction network.
This method was applied to develop pyrolysis models for slash pine (soft wood), eucalyptus (hard wood) and corn stover. Our results showed that a network of first-order parallel and sequential reactions with as many as five solid pseudo-components must be used to accurately describe the pyrolysis dynamics of these biomass feedstocks. The predicted compositions of biochars exhibit the transition continuum described in the literature (Keiluweit and coworkers, ibid.) Starting with a mixture of pseudo-components and intermediates at low pyrolysis temperatures, we move to biochars containing increasing amounts of highly structured char C.
The kinetic parameters of the pyrolysis reaction networks were also estimated using experimental data obtained with different particle sizes. We found that large particle sizes (>1 mm) shift the observed decomposition temperatures to higher values and, thus, strongly influence the kinetics of the endothermic and exothermic pyrolysis reactions. The implications of these findings for designing pyrolysis reactors will finally be discussed.