This document summarizes research on using model predictive control (MPC) for optimizing the operation of large-scale drinking water networks. Key points:
- MPC aims to reduce energy costs while meeting demand and respecting constraints, using forecasts of water demand and energy prices.
- Demand is forecasted using SARIMA, BATS and RBF-SVM models, with RBF-SVM achieving the best accuracy.
- A hydraulic model is developed to predict network state based on inputs, disturbances, and constraints.
- MPC optimizes pumping over a horizon while respecting constraints, using demand forecasts to anticipate future needs.
- Simulation results on a real network show MPC achieving low costs while