Catalytic hydrothermal liquefaction of spent coffee grounds: optimization of process parameters for production of biocrude oil through combined RSM and ANN based methodology

Abstract

In the present study, a central composite design (CCD) approach within the framework of response surface methodology (RSM) was utilized to model and optimize the hydrothermal liquefaction (HTL) of spent coffee grounds (SCGs). The experimental design evaluated the interactive and individual effect of key process parameters such as temperature (270–330 °C), reaction time (20–60 min), and feed concentration (5–15%) on the yield of bio-crude oil. The developed statistical model successfully predicted optimal operating conditions for maximizing biocrude production. The developed model exhibited strong statistical validity, with R² value of 0.95 and minimal prediction error (∼0.12). The optimal HTL conditions were identified at 327.3 °C, a reaction time of 48 min, and a feed concentration of 5 wt%, resulting in maximum bio-crude yield of 34.0 wt%. The close agreement between the experimentally observed data and the values predicted by both RSM and artificial neural network (ANN) models validates the predictive reliability of both modeling approaches. Subsequently, series of industrial catalysts, including potassium hydroxide (KOH) and iron (Fe) were introduced under the optimal HTL conditions to enhance biocrude quality. KOH demonstrated the highest catalytic effect, achieving a maximum biocrude yield of 37.3 wt% coupled with a significantly reduced oxygen content of 10.01 wt%, indicating improved fuel properties. Product characterization using gas chromatography-mass spectroscopy (GC-MS) revealed that the biocrude consist of fatty acids, fatty acid amides, complex amides, and oximes etc. Spectroscopic and compositional analysis confirmed the presence of various organic functional groups and aromatic structures. These findings suggest the potential for refining biocrude into transportation-grade fuels and extracting high-value chemicals.