Machine Learning for Prediction of Inorganic Nitrogenous Compounds Concentration in Swine Wastewater Treatment Plant Effluent

Abstract

In Japan, sum total value of NO₂^－-N concentration, NO₃^－-N concentration, and 0.4 times of NH₄⁺-N concentration in effluent of wastewater treatment plant is regulated by Water Pollution Control Low. Though complying the standard value is important task of swine farms, control of the concentration is generally difficult for farmers. This study attempted to obtain prediction models of “NH₄⁺-N×0.4+NO₂^－-N+NO₃^－-N” fluctuation by machine learning method in two swine wastewater treatment plants. Decision tree model which is one of the methods of machine learning were employed to describe a correlation between operational conditions (aeration tank water temperature, pH, and MLSS as inputs) and the concentration as an output, using data collected over about 3 years. Plant operation guidelines were provided by the decision tree models.