It is envisioned that future Commercial and Institutional (CI) buildings will be installed with a multitude of sensors that will provide water quality measures in real-time. Since water quality does not change instantly, one needs to predict how water quality will change over time and space, and proactively control water quality using chemical additives, or removing contaminants from water, to keep its quality within the regulated safe range for drinking. Therefore, the control systems must first develop water quality functions that can be used in predictive models and that can subsequently be used in a feedback control system to make optimal control decisions on additives, filtering, flushing, pressure and temperature adjustment, and other available control actions. This paper focuses on the development of these predictive models. Water flow including its physical properties (temperature, pressure, etc.), its chemistry (chlorine, pH levels, etc.), and its contaminants are simulated using the easily accessible water network modeling software EPANET. Resultant spatial-temporal water quality data from the simulation model was used to develop macro water quality functions using nonlinear regression and machine learning methods. The computational evaluation using supervised learning shows that these models can predict the water quality well