Politecnico di Torino | Servizi per la didattica

KEYWORD

Digital Twins and Reinforcement Learning for Energy Distribution Optimization

keywords DIGITAL TWIN, ENERGY DISTRIBUTION, OPTIMIZATION, INTELLIGENT TRANSPORTATION SYSTEMS, REINFORCEMENT LEARNING

Reference persons GIUSEPPE CARLO CALAFIORE

Thesis type RESEARCH ORIENTED

Description Efficient distribution of compressed natural gas (CNG) to customers is a complex challenge faced by gas companies, involving factors such as varying customer demands, transportation constraints, and uncertainties. This research proposes a novel framework that integrates digital twin technology, optimization models, and reinforcement learning techniques to enhance decision-making processes in CNG distribution networks. The key objectives are to develop a digital twin system for the CNG distribution network in the Southeast region, formulate an optimization model for vehicle scheduling and dispatching, and apply reinforcement learning algorithms to continuously improve distribution decisions. The digital twin system will enable simulations and scenario analyses, while the optimization model will identify optimal scheduling strategies under various constraints, such as plant capacity, vehicle availability, and time windows. Reinforcement learning techniques will be employed to learn from real-world data or dispatcher feedback, leading to adaptable and improved decision-making. This integrated framework has the potential to maximize profitability, ensure customer satisfaction, and efficiently manage resources in the natural gas industry, contributing to enhanced operational efficiency and sustainable energy distribution.