Data-driven decision making to improve operational efficiencies in leakage control operations

Abstract

Water scarcity poses a critical challenge in Malta, where limited freshwater resources and the high costs of desalination underscore the need for efficient water management strategies. This project explores the potential of data-driven decision-making (DDDM) and machine learning (ML) to enhance leakage control operations within the Water Services Corporation (WSC). By integrating advanced analytics and ML techniques, this research aims to optimize the identification, prioritisation, and management of leaks, contributing to more effective operational practices. Adopting a predominantly quantitative approach, utilizing extensive flow data and district metered area (DMA) performance characteristic, the project developed classification ML models, including Random Forest and Multi-Layer Perceptron, to assess high-leakage zones. Testing showed that Random Forest models were particularly adequate for the leakage classification tasks adopted in this project, achieving excellent accuracy. Muti-Criteria Decision Analysis (MCDA) methods were also utilised to develop a data-driven prioritisation score model for prioritising DMAs more effectively. The findings reveal that data-driven approaches have significant potential at improving the precision and efficiency of leakage management, enabling more targeted interventions and reducing water loss. The results highlight that leveraging big data and ML models can provide actionable insights, streamline operations, contributing to sustainable water management. The proposed data-driven framework offers a relatively simple and easy to implement methodology for more effective, data-driven decisions during leakage management operations. Furthermore, this established a baseline for potential future works in applying such data-driven approaches to on-the-ground leakage operations both in the Maltese, and worldwide context.