Modeling Wastewater Treatment Performance of a Vegetated Constructed Wetland Using Neural Network Approach

Abstract

A multi-layer perceptron neural network model with Levenberg-Marquardt algorithm (MLPLM) was developed based on the performance and operation data of a research worker named Marahatta on a full-scale vegetated submerged wetland system (VSB) operated for a 5 year period. Influent chemical oxygen demand (COD_inf), volatile suspended solids (VSS_inf), total solids (TS_inf) and temperature (T) were determined as the inputs of the model, whereas the output variables were one of the following; (i) effluent chemical oxygen demand (COD_eff), (ii) total solids (TS_eff) and (iii) volatile suspended solids (VSS_eff). Multi-linear regression (MLR) and multi non-linear regression (MNLR) techniques were also used for data analysis to compare the prediction capability. Four criteria used for a statistical comparison were the following: mean square error (MSE), mean absolute error (MAE), mean absolute relative error (MARE) and determination coefficient (R²). The results showed that MLP-LM approach predicted the performance of the constructed wetland system than the MLR and MNLR techniques.