Current students

GERMANO JACOPO

Cycle: XXXVII

Section: Computer Science and Engineering
Advisor: GATTI NICOLA
Tutor: ALIPPI CESARE

Major Research topic:
AI-Powered Regulation of Smart Gas Distribution Networks

Abstract:

For the first time in the history of Mankind, our survival as a species, as well as the equilibrium of the world ecosystem as we know it, is threatened by global-scale phenomena such as climate change, ocean acidification, mass deforestation, depletion of essential natural resources (freshwater, coal, oil, natural gas, fish), air pollution. All of these phenomena are either direct or indirect consequences of human presence and activities. The main goal of this research is to employ techniques from the domains of Machine Learning and Algorithmic Game Theory to help address these environmental issues, by applying existing solutions and by contributing to the development of new ones. As a first case study, we start with the AI-powered regulation of the operational pressures in a natural gas distribution network. Methane gas has a greenhouse effect on the atmosphere which is 70 times that of carbon dioxide. To guarantee enough gas supply despite the time-varying demand, operators often regulate the pressures to much higher values than necessary, leading to increased losses in the network. This high regulation is mainly due to two reasons: the aforementioned time-variability of the demand, which would be too much labor-intensive to forecast through human expertise alone, and the cost of setting different pressure values, which often requires a technician to physically reach the device. With the introduction of smart regulation devices which can set the pressure remotely and constantly monitor the demand, we can now aim at optimizing the operational pressures, to reduce the losses and therefore the impact of the network on the global greenhouse effect. Our goal in this scenario is to develop a data-driven AI algorithm capable of controlling the network pressures, providing service quality guarantees while reducing the emissions.