Economic Model Predictive Control of Water Distribution Systems with Accelerated Optimization Algorithm
Saskia Putri, Faegheh Moazeni, Javad Khazaei
TL;DR
This work addresses the high computational burden of economic MPC in large-scale water distribution systems by introducing a least-restrictive interpolated move-blocking (IDIB) strategy that reduces decision variables while preserving feasibility. The method uses piecewise linear delta-input blocking with predefined block lengths and positions, coupled with time-varying penalties to balance energy cost, tank-level security, and actuator smoothness. Evaluations on an aggregated WDS show about an 80% reduction in online computation and $MAPE<10\%$ across variables, with trajectories closely matching full-DoF MPC. The approach enables real-time, cost-effective demand-driven control for complex urban water networks, offering practical impact for operators and planners.
Abstract
Model predictive control (MPC) has emerged as an effective strategy for water distribution systems (WDSs) management. However, it is hampered by the computational burden for large-scale WDSs due to the combinatorial growth of possible control actions that must be evaluated at each time step. Therefore, a fast computation algorithm to implement MPC in WDSs can be obtained using a move-blocking approach that simplifies control decisions while ensuring solution feasibility. This paper introduces a least-restrictive move-blocking that interpolates the blocked control rate of change, aiming at balancing computational efficiency with operational effectiveness. The proposed control strategy is demonstrated on aggregated WDSs, encompassing multiple hydraulic elements. This implementation is incorporated into a multi-objective optimization framework that concurrently optimizes water level security of the storage tanks, smoothness of the control actions, and cost-effective objectives. A fair comparison between the proposed approach with the non-blocking Economic MPC is provided.