Abstract
Real-time and offline monitoring, control and optimization of significant variables of bio-digester plants is crucial for optimal yield and maximum recovery of bioenergy at an industrial scale. Methane is an energy carrier and a critical component in the total biogas generated in a digestion process, thus necessitating more attention on the fraction of methane in the biogas yield. However, most previous studies in literature had focused on the volumetric yield of methane with little or no attention given to the fractional composition of methane in the total biogas produced. The deficiency of the classical technique in controlling the process parameters for optimal yield has motivated the need for machine learning-based techniques for modelling the methane fraction of biogas in a large-scale plant. In this study, a fuzzy c-mean (FCM)-clustered adaptive neuro-fuzzy inference system (ANFIS) was developed to model the methane fraction of biogas in an industrial-scale plant. The FCM clustering technique was preferred owing to its computational speed boost capability. The model was simulated using the control parameters of the algorithms while the optimal model was selected after testing their performance using relevant statistical metrics. The best model was obtained with ANFIS-FCM model with 8 clusters giving Root Mean Square Error (RMSE), Mean Absolute Deviation (MAD), Average Absolute Percentage Relative Error (AAPRE), Relative Mean Bias Error (rMBE) and correlation coefficient (R2) values of 3.156, 2.236, 3.015, 0.306, 0.978 respectively at the training phase and 4.936, 3.245, 3.456. 0.306, 0.956 respectively at the testing phase. The statistical metrics values obtained implied that FCM-ANFIS is a satisfactory model to predict methane fraction successfully.