The continuous delivery of safe drinking water to any water consumer in a region is fundamental to society. Water Supply Systems (WSS) are responsible for this task, which includes maintaining the infrastructure’s physical integrity and respecting water demands. This work showcases the implementation of a Smart Predictive Digital Twin in the management of WSS. A Smart Predictive Digital Twin is a novel iteration of the Digital Twin upgrading it with predictive and decision-making capabilities. Although there is a scarcity of Digital Twins applied to WSS in the literature, there are some instances of these systems being applied for leak detection [1], pump operation [2], and network design [3]. This study enhances an existing Smart Predictive Digital Twin within a multiservice architecture by focusing on the Digital Model’s capability in mimicking WSS behaviors. Techniques derived from Physics-Informed ML [4] and Transfer Learning [5] are used to develop the ML model. These techniques leverage synthetic data generated by hydraulic simulators, such as EPANET [6], and real-world data. Although hydraulic simulators follows the hydraulic laws inherently in the water network, the ML model relies on the quality and quantity of the provided data (i.e., real-world data). This can cause discrepancies between the hydraulic simulator’s prediction and observed behaviors. Subsequently, depending on the success of calibration, the hydraulic simulator may model some nodes/links more accurately than others. This work considers this by adequately balancing synthetic knowledge and real-world data during training of the ML model. A comparison study is conducted on the proposed methodologies using both synthetic and real-world datasets from a WSS case-study in Portugal. The proposed methodologies show an increase in accuracy when validated on the real-world test dataset. As expected, it was observed that the feature specific accuracy is dependent on the methodology and the extent of the physics-informed and data-driven contributions during learning. Implementing a multiservice Smart predictive Digital Twin to manage a WSS is a promising approach to handling existing multifaceted challenges. Developing a physics-informed Digital Model improves the predictive capability of the overall system. By improving the modeling of hydraulic behavior, it enables the usage of advanced control algorithms.