Performance prediction and upgrading of electroanaerobic baffled reactor using neural-fuzzy method

Abstract

Considering that awareness of the effects of changes in the important operating variables of a wastewater treatment reactor is essential for its designing and maximizing efficiency, this study is focused on prediction and optimization of the performance of the anaerobic baffled reactor upgraded by an electrolysis process (EABR) using an adaptive neural-fuzzy inference system (ANFIS). A semi-industrial pilot of an anaerobic baffled reactor with a total volume of 72 L was integrated with an electrolysis system, and its operation was monitored for three months, during which 42 datasets were collected. Based on the measurements performed during this time, simulation, prediction, and optimization of the desired output variables, including methane yield, effluent COD, and pH value of the fifth chamber, were done. The input factors of the ANFIS model included the pH value of the second chamber, HRT, current density, influent COD, and voltage. Different ANFIS models were evaluated from the point of view of the cluster radius. The results indicated that the radii of 0.25, 0.15, and 0.85 have the best performance to simulate and predict the methane yield, effluent COD, and pH value of the fifth chamber, respectively, with the prediction, mean absolute percentage error (MAPE) values of 2.1276%, 2.6178%, and 0.2217% and the prediction correlation coefficients of 0.9801, 0.9933, and 0.9699. Finally, using the 10 three-dimensional response surface graphs obtained for each output variable, the optimization of the reactor's important operating variables was conducted to achieve the maximum treatment efficiency. This study showed that the ANFIS model is a proper tool for simulating, predicting, and optimizing EABR and managing the intended operating conditions.