Comparative evaluation of Artificial intelligence based models and kinetic studies in the prediction of biogas from anaerobic digestion of MSW

Abstract

The generation of biogas from the decomposition of MSW increases greenhouse gas emissions, contributing to global warming. However, the produced biogas can be effectively captured for conversion into a renewable energy resource. This study demonstrates the utility of artificial neural networks (ANN), adaptive network-fuzzy inference systems (ANFIS), and kinetic models in predicting biogas production from the anaerobic digestion of MSW. The digestion time, volatile solids, pH, temperature, and moisture content were selected as the independent variables. Furthermore, kinetic studies were conducted to predict the biogas yield using First-order, modified Gompertz, Logistic, and Transference models. To assess the efficacy of these models, their performance was evaluated using the coefficient of determination (R2) and root mean square error (RMSE). Upon comparing the results, it was evident that for set IX under 100 % field capacity (FC), the ANFIS network exhibited superior performance with RMSE and R2 values of 0.670 and 0.999, respectively. In comparison, the ANN and modified Gompertz models yielded RMSE values of 0.863 and 21.705, and R2 values of 0.999 and 0.998, respectively. While the coefficient of determination (R2 ≥ 0.99) suggested satisfactory agreement between all the models and experimental data, the ANFIS and ANN models achieved notably lower RMSE values, demonstrating their competitive performance.