Assessment of the differential evolution method for control of multiple-sources water supply system – case study

Abstract

This study demonstrates the use of the differential evolution method for optimizing multi-source water supply systems. The objective was to identify a water-supply scenario that minimizes the total water-supply costs in multiple sources network. The analysis assumed varying water-production costs at each source, with control implemented through the adjustment of valve settings at the network inlets. For the research, open-source tools such as EPANET, the WNTR library, and Python’s scipy.optimize package, were utilized. The analyses were conducted on two water distribution networks. The first was a simplified network with three water sources, designed to validate the simulation environment and perform initial assessments of the optimization algorithms, whereas the second was a complex model representing a real water distribution network. Analyses were conducted for the latter assuming three, four, five, and six sources. For both networks, multiple pricing scenarios were evaluated to explore how different conditions impact the performance of the algorithm. The results show that near-optimal solutions can be achieved with significantly reduced computation times. The differences in objective function values between the brute force and differential evolution (DE) methods were no greater than 0.2%, while the computation time for DE was substantially shorter. These findings highlight DE scalability, efficiency, and potential for real-world applications, promoting cost-effective and accessible optimization in engineering practice.