Obtaining Key Parameters and Working Conditions of Wastewater Biological Nutrient Removal by Means of Artificial Intelligence Tools

The oxidation-reduction potential (ORP) and the dissolved oxygen (DO) have been monitored in a municipal wastewater treatment plant (WWTP). Three thousand two hundred aeration-non-aeration cycles were recorded. They were analyzed by defining 16 parameters to characterize each one of them. The vectors so obtained were treated with the box-plot tool to reject those with outliers (abnormally high or low values). The remaining data were processed by a neural network (self-organizing map: SOM) in order to classify them into classes and to obtain relations between parameters to identify those more representative of the system dynamics. They were: the oxygen uptake rate (OUR), the oxygen rise average slope (ORAS), and the oxidation-reduction potential arrow (ORParrow, the maximum distance between the ORP curve and its linearization). Finally, the classes obtained from SOM were grouped into four macro-classes by means of the K-means algorithm in order to define four operation states related to seasonal and load characteristics, which may be taken into account, along with the key parameters, in the WWTP management with the aim of improving the nutrient removal performance by adapting their controllers to seasonal and load variations.