Photovoltaic self-consumption optimization for Home Microgrid: A Deep Reinforcement Learning approach

Mohamed Saâd EL HARRAB (1) , Michel Nakhla (1)

1 CGS i3 - Centre de Gestion Scientifique i3

Increasing penetration of renewable energy sources (PV, Wind) due to environmental constraints, impose several technical challenges to power system operation. The fluctuating and intermittent nature of wind and solar energy requires constant supply-demand balance for electric grid stability purposes. Self-consumption is a regulatory framework intended to promote local consumption over export. Thus, self-consumption will raise the profit of PV electricity from grid-connected residential systems and lower the stress on the electricity distribution grid. This work presents a novel Deep Reinforcement Learning (DRL) Based Energy Management System (EMS) to control a Home Microgrid system powered by renewable energy sources (PV arrays) and equipped with an energy storage system. An optimal energy scheduling is carried out to maximize the benefits of available renewable resources through self-consumption. A DRL approach is used to make optimal decisions and generate the optimal management strategies.

Mots clés: Deep Reinforcement Learning; Prosumer; Self-consumption; Microgrid; Smartgrid; Optimal scheduling

Domaines: Intelligence artificielle [cs.AI]; Optimisation et contrôle [math.OC]; Gestion et management

Mohamed Saâd EL HARRAB, Michel Nakhla. Photovoltaic self-consumption optimization for Home Microgrid: A Deep Reinforcement Learning approach. EURO 2022, 32nd EURO Conference, Association of European Operational Research Societies, Jul 2022, Espoo, Finland. ⟨hal-03746179⟩