Evaluation of the predictive capability of Response surface methodology (RSM) and Adaptive neuro-fuzzy inference system (ANFIS) in estimating biogas yield from poultry droppings, pig manure & brewery spent grain using enzyme amylase

Abstract

Abstract One of the biggest obstacles investigators face is predicting biogas yield, which is an enormous challenge because study in the field of simulation and optimization of biogas yield is still restricted, particularly with the adaptive neuro-fuzzy inference system (ANFIS). This study aims to develop and compare models that best predict the oxygen-independent decomposition of poultry droppings, pig manure and brewery spent grain for biogas production with the influence of enzyme amylase. The Response surface modelling (RSM) and a hybrid algorithm, Adaptive Neuro Fuzzy Inference System (ANFIS), was evaluated by considering temperature, residence duration, enzyme concentration, and pH as independent variables, while biogas generation served as the response or output factor. Relevant statistical metrics like AAD, MAE, RMSE and \({R}^{2}\) were applied to relate the adequacy of the two models. The performance of both RSM and ANFIS were compared based on the performance metrics. The calculated value of the coefficient of determination (\({R}^{2})\) for the biogas yield by implementing RSM using the enzyme amylase gave (\({R}_{PD}^{2}=0.9974, {R}_{PM}^{2}=0.9910 \& {R}_{BSG}^{2}=0.9975)\). This was compared with ANFIS results with an \({R}^{2}\) value of \(( {R}_{PD}^{2}=0.9986, {{R}_{PM}^{2}=0.9985 \& R}_{BSG}^{2}=0.9985).\) Analysis of the RSM and ANFIS results shows that the ANFIS prediction result is statistically marginal and presents the hybrid algorithm as a better prediction and optimization tool.