Genetic-Algorithm Optimization of PRV Placement for DeePC-Based Pressure Control in Water Distribution Networks

Pressure management in water distribution systems depends not only on valve operation but also on the placement and number of pressure-reducing valves (PRVs). This study develops a controller-in-the-loop framework coupling a Genetic Algorithm (GA) with Data-Enabled Predictive Control (DeePC) to optimize internal PRV placement according to closed-loop performance. The GA searches the feasible space of valve-location configurations, and each candidate is evaluated through hydraulic simulation and closed-loop DeePC control according to four performance criteria: tracking accuracy, spatial pressure variability, pressure-bound violations, and valve actuation effort. The framework was tested on the Fossolo and Modena networks. In Fossolo, the GA identified a three-PRV configuration that achieved both the lowest mean absolute error and the best combined-objective score. In Modena, configurations with two to six added internal PRVs were compared using a consistently normalized global objective. The two-PRV configuration achieved the best overall score of 0.547, although the lowest tracking error and spatial variability were obtained with four and five PRVs, respectively. These results show that adding more controllable valves does not necessarily improve overall performance and that PRV placement should be assessed jointly with controller behavior and multiple operational criteria. The proposed GA-DeePC framework demonstrates a proof-of-concept controller-in-the-loop approach for integrating PRV placement with closed-loop pressure-control performance.