Urban Water Quality Monitoring and Analysis Using Event Detection System and Machine Learning Methods

Abstract

Water quality is defined as the measure of physical, chemical, and biological characteristics of water. Monitoring water quality is a growing challenge because of accidental or intentional spills of industrial, domestic, and agricultural wastes into surface water. Conventional methods used for measuring water quality parameters are time-consuming and expensive, making real-time contamination detection difficult. Advanced monitoring technology can be employed for real-time monitoring, providing a reliable and cost-effective solution to water management. CANARY event detection system (EDS) has been used in water distribution networks and wastewater treatment plants for detecting anomalous water quality events and proved to be an effective alternative to manual laboratory analysis. This dissertation is directed towards analyzing different methods for real-time water quality monitoring, identifying quality trends, and predicting water quality using CANARY and machine learning (ML) techniques. The research provides an insight into the effectiveness of CANARY and ML algorithms for surface water quality monitoring. Considering the effectiveness of CANARY in real-time contamination event detection, this study evaluated the application of the EDS to river water quality analysis and beach bacterial contamination monitoring. For more efficient water quality data management and pollution control, ML models have been developed for water quality monitoring and prediction of different water quality variables, including biochemical oxygen demand (BOD5), total organic carbon (TOC), and Escherichia coli (E. coli) bacteria. The significance of this dissertation is the first successful application of CANARY to natural source water and the development of novel ensemble-hybrid ML models in predicting surface water quality.