Comparison of conventional and machine learning models for kinetic modelling of biomethane production from pretreated tomato plant residues

Tomato plant residues ( Solanum lycopersicum L. ) lack sustainable applications as abundant lignocellulosic biomass after harvest. These residues can be utilized as substrates in anaerobic digestion for biomethane production, generating energy and reducing waste. The purpose of this study was to investigate the sustainable utilization of tomato plant residues for biomethane production at varying conditions and to model biological kinetics. The study aimed to evaluate the effects of varying substrate/inoculum ratios, sulfuric acid pretreatment concentrations, and yeast ( Saccharomyces cerevisiae ) addition on biogas and biomethane yields under mesophilic conditions (37°C). Maximum biogas and biomethane yields in the studied range were obtained when the substrate/inoculum ratio was 3 (g substrate/g inoculum), the sulfuric acid concentration used for residue pretreatment was 2 %v/v, and the substrate/yeast ratio was 10 (g substrate/g yeast). The yeast ratio of 10 increased the cumulative biogas and biomethane production by 96.5 and 128.9 %, respectively. Conventional models (Modified Gompertz, Cone, First-order, Logistic) and Machine Learning models (Support Vector Machine and Neural Network) were compared for biological kinetics. Machine Learning models were also observed to give good fitting results similar to conventional models. Results suggest that Machine Learning models (RMSE: 2.5833–12.0500) are reliable methods like conventional kinetic models (RMSE: 2.1796–13.4880) for forecasting biomethane production in anaerobic digestion processes and Machine Learning models can be applied without needing prior understanding of biomethane production kinetics. • Biogas and biomethane production from tomato plant residues with waste sludge. • Models for forecasting biomethane production in anaerobic digestion processes. • Machine Learning models are compatible with the conventional biogas kinetic models. • Support Vector Machine and Neural Network models were suitable for modeling.