A data mining approach to the prediction of food-to-mass ratio and mixed liquor suspended solids

Abstract

This paper presents methodology for constructing a statistical model to forecast food-to-mass ratio (F/M). In the model, wastewater inflow (Q), biochemical oxygen demand (BOD5) and mixed liquor suspended solids (MLSS) were modelled separately using artificial neural networks (ANN) and multivariate adaptive regression splines (MARS). To compute the value of MLSS, the quality indicators of influent wastewater and the operational parameters of the bioreactor were used. It was examined whether it is possible to predict wastewater quality indicators that determine the values of F/M and MLSS on the basis of the wastewater inflow to the treatment plant. Computations performed demonstrated that ANN predictions of MLSS and F/M showed smaller errors than those obtained using the MARS method. Moreover, all developed models of wastewater quality indicators were considered as satisfactory.