Biomass Hydrothermal Carbonization: Markov-Chain Monte Carlo Data Analysis and Modeling

Abstract

This paper introduces Bayesian statistical methods for studying the kinetics of biomass hydrothermal carbonization. Two simple, specially developed computer programs implement Markov-chain Monte Carlo methods to illustrate these techniques' potential, long since established in other areas of chemical reaction engineering. A range of experimental data, both from this study and the literature, test the soundness of a Bayesian approach to modeling biomass hydrothermal carbonization kinetics. The first program carries out parameter estimations and performs better or equal than the traditional deterministic methods (R2 as high as 0.9998). For three out of the 22 datasets, the program detected the global minima of the parameter space, while the deterministic least-square found local values. The second program uses Gillespie's algorithm for the statistical simulation of the reactions occurring in hydrothermal carbonization. Comparing six basic kinetic models with literature data tested the stochastic simulation as a tool for assessing biomass conversion reaction networks rapidly. Among the simple models discussed, reaction scheme 3 fitted better to the experimental data (R2 > 0.999). The proposed approach is worth extending to more complex, time-consuming computer models and could support other techniques for studying hydrothermal conversions.