Modern Predictive Modelling of Energy Consumption and Nitrogen Content in Wastewater Management

Abstract

Water and energy resources play a vital role in daily life, resulting in increased wastewater production, emphasizing the relevance of wastewater treatment plants, as well as the need to control the plant's energy consumption. The objective of this study was to analyse and predict the total energy consumption in wastewater management followed up with a detailed analysis and prediction of total nitrogen content that can be extracted from it for various applications. We had considered data from the Eastern Wastewater Treatment Plant in Melbourne. To obtain quantified relationships of wastewater parameters with energy consumption and total nitrogen, multiple predictive machine learning algorithms such as regression, support vector regression and Ensemble model had been implemented. Data pre-processing and feature selection methods based on Principle Component Analysis were used to curate four-input parameters in the prediction of total nitrogen and six parameters in the prediction of total energy consumption. A Correlation Matrix was plotted and analysed, which resulted in the selection of a three-input parameter model for predicting energy consumption and two-input parameters for predicting total nitrogen. The predictive models were evaluated based on Root Mean Square Error, Mean square error and Mean absolute error. It was discovered that the support vector regression model with Radial Basis Function kernel provided significant performance for both energy consumption and total nitrogen prediction. Polynomial regression models, in addition to the support vector regression model with the Radial Basis Function kernel, would be a good choice for energy consumption prediction.