A Federated Learning Approach for Water Distribution Networks Monitoring

Nowadays, people's demand for water is growing as well as in public, commercial, and industrial sectors. However, the limited character for water resources presents a crucial obstacle to satisfying needs for continued human development. The control and provision of potable water are therefore the most vital challenges for the water supply system. It must use water resources in an efficient manner, and fulfill both quality and quantity demands. For this purpose, the current water supply system uses smart infrastructure that gathers, processes, stores, and delivers water from water sources to users. It is done in a very complex environment with ever-increasing demand, and often conflicting services to deliver. In this work, we developed a Federated Learning (FL) model, which is come in the form of a machine learning setting where clients collaboratively train under the orchestration of a central server while keeping the training data decentralized. The FL embodies the principles of focused data collection and minimization, it can mitigate many of the systemic privacy risks and costs resulting from traditional, centralized machine learning and data science approaches. Implementing on the central server, the FL will iterate on the gateway's learning models using the federated averaging algorithm. Our approach's capacity is shown by addressing reel instances in a real-world issue in which the proposed method finds much better results and a significant improvement in performance compared to the standard approaches while achieving our goal and suggesting a new interesting direction for research.