Application of an optimal predictive controller for a small water distribution network in Luxembourg

This paper deals with predictive control applied to the management of water storage in a small water distribution network. This online optimisation-based strategy is computed iteratively by solving a set of mathematical equations which describe the operative goals, in a given time horizon, and uses a representative model for the network dynamics as well as a demand forecast. The approach has been tested on a simulator developed for a four-reservoir water network of a commune of Luxembourg. Mathematical optimisation of the water distribution network is defined to account for all the requirements put forward by the authorities (both the commune and the regional drinking water provider as well as the national water agency), without ignoring the operating and physical constraints of the network. Based on realistic consumption scenarios, the starting situation (tanks always completely filled) and different control strategies (proportional integral derivative or PID level regulation or global predictive control) have been compared and the results are discussed. Moreover, the question of whether to integrate natural ventilation has been explored. Global predictive control leads to improved management in comparison to PID and mechanical control. Further work is needed to evaluate performance when goals, such as tank aeration, are to be met.