In this paper, two tuning strategies for a multi-objective predictive controller applied to a drinking water network (DWN) are proposed. A control-oriented DWN model is briefly reviewed, together with its management objectives. A comparison of methods to explore the Pareto front of the multi-objective optimisation (MOO) problem behind the predictive controller is presented with an effective normalisation method for the model predictive control (MPC) objectives. The proposed tuning strategies, applied to a real-life case study, are compared. Finally, simulation results show that the proposed MPC tuning strategies outperform the baseline results.

In this paper, the tuning of economic model predictive control (EMPC) applied to drinking water transport networks (DWTNs) is addressed using multi-objective optimization approaches. The tuning strategies are based on Pareto front calculations of the underlying multi-objective problem. This feature represents an improvement with respect to the standard EMPC approach for weight tuning based on trial and error. Different multi-objective optimization methods with corresponding normalization approaches of the controller objectives are first studied to explore the dynamic nature of the Pareto fronts. An automated decision-making strategy is proposed to select the preferred controller parameters as a function of different disturbance values. The tuning requires an offline training phase and an online application phase. During the offline phase, the controller parameters are selected for different disturbances using the decision-making strategy. During the online phase, two approaches are evaluated: (i) exploiting the controller parameters with the highest frequency in the resulting histogram or (ii) using a regression model between the controller parameters and the disturbances. The proposed tuning strategies are applied to a real-life simulation case study based on the Barcelona DWTN. The simulation results show that the proposed tuning strategies outperform the baseline results by exploiting the periodicity of the water demands profile.

No abstract