Transient Pressure Estimation Using Data-Driven Models: An Approach Based on Ensemble Trees

The operation of control devices in a water distribution network establishes a transient flow that leads to pressure oscillation during a certain time. Depending on the pressure wave’s amplitude, infrastructure can be exposed to dangerous pressure levels, collapsing or breaking the pipes. Predicting the maximum and minimum head pressure values caused by an operation can assist in the design of safety devices and increase the lifespan of the distribution network. Nevertheless, the calculation of the pressure wave magnitude based on hydraulic equations can be time consuming, reducing the possibility of an optimal approach for designing a protection system. In this sense, this study presents a data-driven model for predicting pressure values caused by transient flows in a water distribution network. The methodology is implemented using an ensemble tree linked to the XGBoost algorithm and it is applied in two case studies.