Physics-Informed Neural Network Modeling and Predictive Control of District Heating Systems

Physics-Informed Neural Network Modeling and Predictive Control of District Heating Systems

2024 | Laura Boca de Giuli, Alessio La Bella, Riccardo Scattolini
This article addresses the data-based modeling and optimal control of district heating systems (DHSs). The authors propose a novel methodology that leverages operational data and physical knowledge to develop accurate and computationally efficient dynamic models of DHSs. The proposed methodology, called physics-informed recurrent neural networks (PI-RNNs), integrates multiple recurrent neural networks (RNNs) and embeds the physical topology of the DHS network in their interconnections. This approach enables faster training procedures and higher modeling accuracy, even with reduced-dimensional models. The developed PI-RNN modeling technique is used to design a nonlinear model predictive control (NMPC) regulation strategy, which optimizes the operation of the DHS by minimizing production costs, increasing system efficiency, and respecting operational constraints. The effectiveness of the proposed methods is demonstrated through simulations on a benchmark DHS, showing promising results in terms of modeling and control performance.This article addresses the data-based modeling and optimal control of district heating systems (DHSs). The authors propose a novel methodology that leverages operational data and physical knowledge to develop accurate and computationally efficient dynamic models of DHSs. The proposed methodology, called physics-informed recurrent neural networks (PI-RNNs), integrates multiple recurrent neural networks (RNNs) and embeds the physical topology of the DHS network in their interconnections. This approach enables faster training procedures and higher modeling accuracy, even with reduced-dimensional models. The developed PI-RNN modeling technique is used to design a nonlinear model predictive control (NMPC) regulation strategy, which optimizes the operation of the DHS by minimizing production costs, increasing system efficiency, and respecting operational constraints. The effectiveness of the proposed methods is demonstrated through simulations on a benchmark DHS, showing promising results in terms of modeling and control performance.
Reach us at info@study.space
[slides] Physics-Informed Neural Network Modeling and Predictive Control of District Heating Systems | StudySpace