Use of artificial neural network for predicting effluent quality parameters and enabling wastewater reuse for climate change resilience – A case from Jordan

Abstract

Jordan is considered an arid to semi-arid country. Water resources in Jordan comprise of conventional resources (ground water and surface water) and non-conventional resources mainly treated wastewater. In this study, artificial neural network (ANN) models were developed to provide predictions for the quality of treated effluent from Wadi Arab wastewater treatment plant (WWTP)-Phase1. This plant serves South-West of Irbid city. The treated effluent of Wadi Arab WWTP is used for irrigation. Four ANN models were developed to predict effluent quality of Wadi Arab WWTP-Phase1, namely BOD5, COD, SS, NH4-N. A systematic approach was employed to find the most appropriate architecture for each ANN model. The influent flow rate (Q), Temperature, pH, BOD5, COD, SS and NH4-N were the input quality parameters for each ANN model. Sensitivity analysis was conducted to uncover the prediction uncertainty of an ANN model to variations in each input parameter. The analyses showed that all four ANN models were highly sensitive to influent pH,while all were slightly sensitive to influent SS. This study found that ANN models were reasonably robust and they could be used to alert plant staff of future problems such as those indicated in the above scenarios. The results showed that ANN models are beneficial to manage the treatment process in Wadi Arab WWTP, to produce effluent quality within the Jordanian Standards for irrigation, and to protect the ecosystem in the Jordan Valley.