Prediction of effluent quality of an anaerobic treatment plant under unsteady state through ANFIS modeling with on-line input variables

A neural fuzzy model based on adaptive network-based fuzzy inference system (ANFIS) was proposed in terms of on-line input variables CH4%, Q(gas), Q(anarecycle), Q(inf-bypass) and Q(inf) to estimate the effluent chemical oxygen demand. CODeff, of a real scale unsteady anaerobic wastewater treatment plant of a sugar factory. Two new variables were added into the input variables matrix of the model; phase vectors of the plant operation and the history of effluent COD values in order to increase the fitness of simulated results. ANFIS was able to estimate the water quality discharge parameter with success for the case when only limited on-line variables were available without requiring the measurement of inlet COD. Acceptable correlation coefficient (0.8354) and root mean square error (0.1247) were found between estimated and measured values of the system output variable, effluent COD, in the case of excluding inlet volumetric flow rate of the wastewater treatment plant from the on-line input variable matrix. The developed ANFIS model may be integrated into an advanced control system for the anaerobic treatment plant using different control strategies with further work. (C) 2008 Elsevier B.V. All rights reserved.