Pressure-dependent leakage modeling and AI-assisted control in a rural water supply network: Varzea da Cobra (Ceará, Brazil) case study

ABSTRACT High levels of water loss due to leakage remain a critical challenge for small rural water supply systems, where technical capacity and financial resources are often limited. This study evaluates pressure management strategies in the Várzea da Cobra distribution network, located in northeastern Brazil, through a simulation-based approach. A hydraulic model incorporating pressure-dependent leakage was calibrated with Minimum Night Flow data to estimate real losses under local operating conditions. Three scenarios were analyzed: no pressure control, installation of a pressure reducing valve (PRV) with a fixed outlet setting and day/night modulation, and a dynamic PRV operated by an AI-assisted controller. Results showed that the fixed PRV strategy reduced daily leakage by approximately 24% without compromising service levels. The AI-assisted PRV achieved comparable leakage reduction while maintaining more stable pressures throughout the day. To facilitate practical application, a Python-based decision-support tool was developed, enabling non-specialist operators to simulate and evaluate control strategies with minimal input data and Excel-based outputs. The findings demonstrate that effective pressure management, whether through simple mechanical solutions or advanced optimization, can significantly reduce leakage and improve reliability in rural water systems. This approach offers a scalable, low-cost pathway for enhancing water supply sustainability in developing regions.