Prediction of Biochemical Oxygen Demand in a Wastewater Treatment Plant by Artificial Neural Networks

Abstract

In this study, output biochemical oxygen demand concentration of Kayseri advanced biological wastewater treatment plant was defined with the daily input data of 2004-2007 belonging to the same facility and this data was estimated rapidly and confidently by training with multi layered artificial neural networks model. In the establishment of the artificial neural networks model temperature, total nitrogen, total phosphorus, suspended solids, chemical oxygen demand and total dissolved solids parameters were used as input while biochemical oxygen demand parameter was used as output. The structure yielding the best result was obtained by training the artificial neural networks structure with 5 inputs, two hidden layers by Levenberg-Marquardt algorithm. In this structure, it was found that mean square error 0.45, mean absolute error 0.445 and R² = 0.915.