Application of artificial neural network models for predicting wastewater treatment plant performance and effluent quality

Abstract

In this study, Artificial Neural Networks (ANN) were employed for monitoring pollution levels in a wastewater treatment plant and predicting the quality of treated effluent. A Multi-Layer Perceptron (MLP) architecture, a class of feedforward neural networks trained via the backpropagation algorithm, was developed to model the behavior of three key water quality indicators: Biochemical Oxygen Demand (BOD), Chemical Oxygen Demand (COD) and Total Suspended Solids (TSS). Separate predictive models were constructed for each parameter. The ANN-MLP model was implemented and evaluated using two distinct configurations with five and ten hidden neurons respectively. Statistical indicators including Mean Absolute Error (MAE), Root Mean Square Error (RMSE) and coefficient of determination R 2 were used to evaluate model predictive accuracy. Results showed that both configurations achieved high predictive performance, with R 2 ranging from 0.80 to 0.97 for the five-neuron model and reaching up to 0.98 for the ten-neuron model. These results highlight the reliability of ANN-MLP model in capturing complex nonlinear relationships in wastewater treatment processes, demonstrating its value as a tool for treated wastewater quality management and control.