Comparative Study of Wastewater Treatment Plant Feature Selection for COD Prediction.

This study is focused on wastewater treatment station instrumentation. It is preliminary work for accomplishing virtual sensors in the future. In this stage, a specific case of correlation between monitored variables is carried out. The main aim is to choose the right and enough variables to predict the chemical oxygen demand (COD) over a wastewater treatment plant. Firstly, four methods for feature selection are implemented with all the monitored variables. After that, three regression techniques are applied to measure the performance of the previous step’s cases. In all cases were obtained an acceptable COD prediction was, which will allow the implementation of virtual sensors in the future with predictably adequate accuracy.