Performance of hydrothermal liquefaction (HTL) of biomass by multivariate data analysis

Hydrothermal liquefaction (HTL) is one of the most promising biomass reforming processes for production of drop-in biofuels. The technique has been under development for a number of years, and yet, due to its complexity, it has always been difficult to generalize information about the optimal conditions. The main issue regards to the limited knowledge available from batch studies evaluating HTL by a finite number of process conditions in certain combinations. In this study, multivariate statistical methods were applied for investigation of HTL data available in the literature. The aim was to determine a set of generally valid rules for prediction of the output from the process (yields and the energy content) on the basis of relatively few process parameters. The results have shown that multivariate data analysis can be used to make predictions about HTL and increase our understanding of the process, despite the fact that the input data constituted a very broad spectrum of values. In general, biomass type and properties were the most significant parameters controlling both the obtained yields and the energy content in the produced biocrude. Regression models calculated for all groups of biomass were relatively poor, due to the lack of common trends. However, a number of statistically sound models were obtained for selected combinations of biomass and responses. The drawn conclusions not only supported the pre-understood axioms of HTL, but also indicated a number of new associations. It was shown that the overall conversion rates are governed by biomass properties and the applied heating velocities, while the amount of homogeneous catalyst and the reaction time control the distribution of the products between the water phase and the biocrude. The energy content in the biocrude produced from lignocellulose was dependent mostly on the biomass content and properties, and not the process conditions.