Institut Catholique d''Arts et Métiers

Nowadays, improving thermal comfort while reducing Heating Ventilation and Air Conditioning (HVAC) energy consumption remains big challenge in smart buildings. This paper proposes Holonic Multi-Agent Systems (HMAS) integrating Predicted Mean Vote (PMV) assessment with Deep Reinforcement Learning (DRL) for decentralized (HVAC) control. The proposed hierarchical architecture enables scalable and adaptive operation across buildings zones. Experimental validations in single and multi-zones simulations demonstrates accordance with ISO7730 and ASHRAE55 comfort standards while achieving energy savings between 20% and 35%. Statical analyses confirms strong agreement with reference PMV models, demonstrating de reliability of the proposed approach for occupant centric smart building management.

The availability and need of electricity vary over the day due to many factors including but not limited to the availability of renewable energy systems, and usage of electricity linked to business and private activities in society. As a result of this, the price of electricity varies substantially throughout the day. With the current trend and policy towards electrification, our households are increasing their energy consumption, it is now common to deliver heating and cooling with heat pumps, resulting in very relevant electricity loads. Considering the energy price variations, there are substantial optimization possibilities associated with the predictive control of these assets. This paper proposes a model predictive control system for the day ahead optimization of energy costs. This controller uses genetic algorithms to shift heating and cooling loads and proves that it is possible to perform such a model and always ensure indoor comfort. In this paper, we focus on defining the hyperparameters under which such a process delivers good economic results without over imposing computation constraints to the optimization system. The model with the optimal hyperparameters can perform a day-ahead optimization in about 1 minute, resulting suitable for field deployment.

In this work, the short-circuit (SC) robustness of 650 V normally-off SP-GaN HEMTs is investigated. The devices under test (DUTs) were subjected to single-event destructive SCs at different drain-source voltage levels. Overall, the DUTs exhibited a significant withstand time of several hundred microseconds. The experimental results first reveal the absence of a critical SC energy threshold, indicating that energy alone does not govern the failure mechanism. Instead, device failure is found to be more thermally driven, occurring once the junction temperature exceeds a specific limit. To support this conclusion, the junction temperature (Tj) evolution during the SC event was estimated using a highly simplified thermal model with a uniform heat dissipation based on the finiteelement method (FEM), implemented in ANSYS APDL and calibrated with device geometries and material parameters extracted from its construction analysis. The simulation results show that all DUTs reached nearly the same Tj threshold at the instant of failure, regardless of the applied VDS. However, due to the simplifying assumptions underlying the FEM simulations presented in this study, the estimated temperatures should be regarded as indicative rather than exact.

This paper presents a real-time cyber–physical framework for hybrid PVT–water systems implemented on the SOLLAB experimental platform. The proposed architecture integrates physics-based thermal modeling with live sensor acquisition and actuator control within a deterministic fixed-step execution environment. A layered structure is introduced, including hardware, sensing, communication, control, data management, and supervision components. Experimental validation demonstrates stable real-time performance with deterministic 1 s sampling, robust Modbus TCP/IP communication, and reliable bidirectional actuation of physical components. The system successfully reproduces key thermal behaviors under real operating conditions while maintaining synchronized interaction between simulation and hardware. The presented framework establishes a deterministic cyber–physical synchronization mechanism between physical assets and their numerical counterpart, forming a validated building-level energy node. This implementation provides a scalable foundation for advanced supervisory control, predictive energy management, and integration into community-level digital energy infrastructures.

This paper investigates the heat transfer enhancement in a two-compartment heat exchanger using nanofluids, employing numerical simulations and deep learning. The study systematically examines the influence of key parameters: Rayleigh number (Ra= 10(6)-10(9)), conductivity ratio (kr=1-15), nanoparticle volume fraction (phi=0-3%), nanofluid temperature (Temp=293-323K), and scaled heat exchanger wall thickness (0.02-0.05). The first compartment features internal heat generation, while the second incorporates baffles and nanofluids to optimize mixing and heat transfer. Computational Fluid Dynamics (CFD) is used to analyze Nusselt number, isotherms, streamlines, velocity vector magnitude, exergy loss, entropy generation, and the Bejan number. Deep learning models are developed to predict and optimize heat transfer performance based on these five input parameters. Results demonstrate that increasing the Rayleigh number and conductivity ratio significantly enhances heat transfer, while nanoparticles higher volume fractions improve performance, albeit with potential viscosity increases. Exergy analysis reveals opportunities for design optimization to minimize entropy generation. The integrated approach of CFD and deep learning provides a powerful tool for optimizing the design and operation of nanofluid-based heat exchangers for improved thermal management in various applications.