Communication Dans Un Congrès Année : 2023

A Machine Learning Framework to Select Surrogate Model of District Heating Network Simulation

Résumé

In order to decarbonize the heating sector, district heating networks have been identified as a viable solution due to their superior energy efficiency, economic viability and environmental advantages over decentralized, individual heating systems. However, they raise several challenges concerning the optimization of their control due to their size and the operational constraints of the energy systems involved. As a result, numerical simulations of these networks are computationally heavy, preventing nearinstantaneous optimal control.

In this work, we present the first building block of an optimization framework for the control of district heating networks using a surrogate model approach based on geometric deep learning. Our iterative approach enables us to identify and select an accurate surrogate model to use in a model predictive control loop. We have trained specific architectures of Graph Neural Networks to emulate a thermo-hydraulic simulator of a district heating network. The surrogate model allows us to drastically reduce simulation time by 1 to 4 orders of magnitude depending on its architecture.

Fichier principal
Vignette du fichier
abstract_boussaid_taha_insa_lyon_france.pdf (216.16 Ko) Télécharger le fichier

Dates et versions

hal-05023109 , version 1 (07-04-2025)

Licence

Identifiants

  • HAL Id : hal-05023109 , version 1

Citer

Taha Boussaid, François Rousset, Vasile-Marian Scuturici, Marc Clausse. A Machine Learning Framework to Select Surrogate Model of District Heating Network Simulation. International Forum on Energy & Informatics, Institute of Science Tokyo, Dec 2023, Bali, Indonesia. ⟨hal-05023109⟩
93 Consultations
167 Téléchargements

Partager

  • More