Prediction of common pollution parameters values in wastewater treatment plants by using artificial neural networks

Abstract

The measurement of the BOD₅ level in wastewater treatment plant (WWTPs) influent takes five days, and using a prediction model to estimate BOD₅ saves time and allows the use of an online control system. This study investigates the application of artificial neural networks (ANNs) in predicting the influent BOD₅ concentration and the (WWTPs) performance in terms of (BOD₅, COD, TSS) effluent concentrations. To determine the best performing ANN network structure and configuration, sensitivity analysis was performed. The results revealed that the ANN model developed to predict the COD concentration performed the best among the three parameters. The best performing ANN models yielded R² values of (0.92, 0.99, 0.94) for the prediction of the BOD₅, COD and TSS effluent concentrations, respectively. The optimal performing models were obtained (four inputs – one output), which indicated that the influent pH, TDS, TSS and BOD₅ greatly affect the (WWTPs) performance as inputs in all models. The developed prediction model for the influent BOD₅ concentration achieved a high accuracy (R²= 0.974) which indicates that the model is viable as an accurate tool for online control and management systems for (WWTPs). Generally, the ANN model provides a simple approach for the prediction of the (WWTPs) complex processes.