Predictive modeling of the primary settling tanks based on artificial neural networks for estimating TSS and COD as typical effluent parameters

Abstract

Abstract A predictive model based on artificial neural networks (ANNs) for modeling primary settling tanks' (PSTs) behavior in wastewater treatment plants was developed in this study. Two separate ANNs were built using input data, raw wastewater characteristics, and operating conditions. The output data from the ANNs consisted of the total suspended solids (TSS) concentration and chemical oxygen demand (COD) as predictions of PSTs’ typical effluent parameters. Data from a large-scale wastewater treatment plant was used to illustrate the applicability of the predictive model proposal. The ANNs model showed a high prediction accuracy during the training phase. Comparisons with available empirical and statistical models suggested that the ANNs model provides accurate estimations. Also, the ANNs were tested using new experimental data to verify their reproducibility under actual operating conditions. The predicted values were calculated with satisfactory results, having an average absolute deviation of <20%. The model could be adapted to any large-scale wastewater plant to monitor and control the operation of primary settling tanks, taking advantage of the ANNs' learning capacity.