Modeling the Biogas and Methane Yield from Anaerobic Digestion of Arachis hypogea Shells with Combined Pretreatment Techniques Using Machine Learning Approaches

This study experimented biogas production from Arachis hypogea shells subjected to combined pretreatments, namely particle size reduction and Fe₃O₄ additives. Further to this, the biogas production based on organic dry matter biogas (oDMBY), fresh mass biogas (FMBY), organic dry matter methane (oDMMY), and fresh mass methane (FMMY) were modeled using machine learning algorithms. A fuzzy c-means (FCM)-clustered Adaptive neuro-fuzzy inference systems (ANFIS) and optimized artificial neural network (ANN) model were developed using significant operating parameters of temperature, retention time, and pretreatment methods as input variables. The maximum daily gas yield of 100.4 lN/kgoDM, 18.4 lN/kgFM, 75.7 lNCH₄oDM, and 16.1 lNCH₄FM were recorded when different particle sizes were combined with Fe₃O₄ additives. Single pretreatment with Fe₃O₄ improves the oDMBY, FMBY, oDMMY, and FMMY by 150, 20.7, 79.39, and 176.19%, while the combination with particle size improves the yields by 256.03, 138.96, 155.74, and 283.33%, respectively. The models’ performances were evaluated using relevant statistical metrics. The ANN model gave Root Mean Square Error (RMSE), Mean Absolute Percentage Error (MAPE), Mean Absolute Deviation (MAD), and Correlation Coefficient (R²) values of 1.9354, 6.3213, 1.0234, and 0.9496. The FCM-ANFIS model with ten clusters outperformed the ANN model with RMSE, MAPE, MAD, and Coefficient (R²) values of 1.2343, 5.2343, 1.2463, and 0.9850. The result shows that the pretreatment applied enhanced the biogas and methane yields of Arachis hypogea shells. This study showed that FCM-clustered ANFIS can predict biogas yield of pretreated Arachis hypogea shells satisfactorily, and it is recommended for other similar studies. Graphical Abstract: [Figure not available: see fulltext.].