Design and validation of a microalgae biorefinery using machine learning-assisted modeling of hydrothermal liquefaction

Abstract

A new microalgae biorefinery system adopts an integration of the hydrothermal liquefaction (HTL) of (wet) microalgae and hydrodeoxygenation (HDO) of biocrude oil to produce green products of green diesel, naphtha, and heavy oil, where the HTL and HDO processes are validated through Aspen Plus® simulator. A hybrid model of HTL of microalgae, which consists of the machine learning-assisted model and the GAMS®-assisted RYield reactor, is used to predict the yield, higher heating value (HHV), and yield component distribution of biocrude oil according to the experimental records of the HTL process from 16 species of microalgae. Three types of machine learning (ML)-assisted modeling of the HTL process are illustrated, where the LSBoost model is validated to be better than other ML-assisted models through the testing procedure for measuring R2 and root mean square error (RMSE).To evaluate the performance of the microalgae biorefinery system, it is noted that the green diesel yield is up to 33 wt% and the energy recovery of HTL of Nannochloropsis sp. is estimated at 75 % higher than around 60 % of the HTL of Trentepohlia.