MODELLING APPROACHES COMPARISON APPLIED TO WASTEWATER TREATMENT PLANT PROJECT

Abstract

ABSTRACT Wastewater treatment system involves several intricate parallel biological processes, which are quite uncertain and difficult to predict. However, for efficient operation of the system, a reliable and straightforward model capable of accurately describing the behaviour of the system is strongly needed. Most of the existing developed models were applied to industrial wastewater treatment plant. This paper investigates the effectiveness in predicting the organic nutrient removal of four (4) different modelling approaches applied to the domestic step feed activated sludge wastewater treatment plant in Malaysia. Computational efficiency and reliability tempted the selection of the techniques such as autoregressive with exogenous input model (ARX), nonlinear autoregressive with exogenous input model (NARX), artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS). Simulation studies revealed that ANFIS model demonstrated better prediction capability compared to the other models in all the considered variables having root mean square error (RMSE) 0.0133 and mean absolute percentage deviation (MAPD) 3.88% for biochemical oxygen demand (BOD), RMSE 0.0478 and MAPD 5.16% for chemical oxygen demand (COD), RMSE 0.0086 and MAPD 3.42% for suspended solids (SS) and RMSE 0.0310 and MAPD 6.40% for ammonium nitrogen NH4-N during validation. This shows that the ANFIS model may serve as a valuable tool for predicting the effluent quality in the wastewater treatment plant.