Using Artificial Neural Networks to Estimate the Return Sludge Rate, A Case Study of Torbat Heydarieh Wastewater Treatment Plant

There are complex and nonlinear causal relationships among the different quality and quantity parameters of wastewater and return activated sludge, which is one of the most important parameters in the operation of activated sludge wastewater treatment plants. On the other hand, Artificial Neural Networks (ANNs) have advantages such as the ability to identify and extract complex and nonlinear causal relations using simple mathematical formulas, high generalizing power, and high speed that make them capable of being used as a powerful software in the operation of wastewater treatment plants. In this study, the application of artificial neural networks (ANNs) for determining the amount of return activated sludge in Torbat Heydarieh wastewater treatment plant is investigated based on one-year inlet flow data. The different parameters involved in this process such as inlet flow and temperature, inlet and outlet total suspended solids, inlet and outlet BOD5 and COD, MLSS, and the amount of return activated sludge were collected and applied to MLP and RBF artificial neural networks (ANNs). Results showed thatMLP is capable of estimating the return activated sludge required in conventional biological wastewater treatment systems such as extended aeration and that its estimation accuracy is above 93%.