Optimization of hydrothermal liquefaction process through machine learning approach: process conditions and oil yield

This study involves an artificial intelligence approach in the optimization of hydrothermal liquefaction (HTL) of biomass feedstock. A Decision Support System (DSS) was developed using machine learning algorithms. Dataset from published work and unpublished dataset from the authors' research team were used in this study. The Pearson correlation matrix was generated for a training dataset of 400. Bio-oil yield showed a high positive correlation of %C, %H of biomass and temperature, and catalysts loading in the HTL process. A high negative correlation was seen among %O, %moisture, and %ash with yield. Weighted ranks were assigned to the influential parameters and predictions were made for optimum HTL process parameters for a testing dataset of 20. To validate the DSS output, laboratory experiments were carried out and the results showed more than 94% accuracy with the predicted data. The machine learning-based optimization method is more suitable for a highly parameter-oriented process like HTL of biomass.