Black-Box Modeling of Water Quality in WDS: A Case Study

Water quality is one of the most important factors in the design of water distribution systems (WDS). To maintain the conditions of water age, residual chlorine, and other quality characteristics, the ability to adjust hydraulic conditions using control elements is necessary. Although there are different methodologies for the analysis of water quality dynamics, the continued development and application of new modeling alternatives and methodologies to more efficiently represent water quality dynamics in WDS are needed to provide efficient tools for managing water quality in real time. The ability to provide water quality information in real-time would provide benefits under normal operating conditions and increase capabilities for fast responses in an unexpected pollution event, whether intentional or unintentional. Therefore, being able to use these new methodologies will enhance the capacity to develop and use adequate models to analyze different alternatives to quickly solve and control problems like the spread of a contaminant in a network. This study presents an initial analysis of water quality input-output data necessary for black-box modeling performed using the EPANET-MATLAB toolkit with respect to an existing WDS network, Bogotá, Colombia. The methodology is based on (1) analysis of model input and output variables, (2) numerical simulation of network water quality in the aforementioned software, and (3) data analytics comparing historical vs. simulated conditions for multiple scenarios. The purpose of this methodology is to estimate the sensitivity of existing models compared with new scenarios, created from different initial conditions for data analysis previous to the use of a black-box model.