2026-04-13T02:43:40Zhttps://recercat.cat/oai/request

oai:recercat.cat:2117/4465182025-11-21T06:31:49Zcom_2072_1033col_2072_452950

2025-11-21T06:31:49Z urn:hdl:2117/446518 Development of an artificial intelligence algorithm for indoor air quality optimization and industrial production management Desarrollo de un algoritmo de inteligencia artificial para optimización de calidad de aire interior y gestión de producción industrial Cebolla Alemany, Joaquim Macarulla Martí, Marcel Viana Rodríguez, Mar Gassó Domingo, Santiago Àrees temàtiques de la UPC::Edificació::Instal·lacions i acondicionament d'edificis::Instal·lacions de ventilació Àrees temàtiques de la UPC::Informàtica::Intel·ligència artificial::Aprenentatge automàtic Nanoparticles Reinforcement learning Indoor air pollution Genetic algorithms Energy Nanoparticles Industry Air extraction Reinforcement learning Genetic algorithm Nanopartícules Aprenentatge per reforç Contaminació de l'ambient interior Algorismes genètics Nanoparticle moNanoparticle modelling allows concentration simulations in industrial settings to estimate the effect of different air extraction strategies in scenarios with nanoparticle emitting processes. Several artificial intelligence-based techniques can evaluate these strategies to find the optimal one. Moreover, they can simultaneously minimise the energy cost of the extraction process by coordinating the industrial activity with the hourly grid energy cost fluctuation. Consequently, two artificial intelligence algorithms are proposed based on genetic algorithms and reinforcement learning. For the first, a population generator manages system’s restrictions based on real operative scenarios and then individuals change through time imitating natural selection, reproduction and mutation processes. For the second, a meta-heuristics policy is designed from state space and actions consisting on different heuristic strategies to explore potential solutions. Preliminary results evaluating the energy cost performance show that both algorithms reach similar solutions, registering the expected population features curve for the genetic algorithm but not illustrating a clear learning curve for the reinforcement learning study.delling allows concentration simulations in industrial settings to estimate the effect of different air extraction strategies in scenarios with nanoparticle emitting processes. Several artificial intelligence-based techniques can evaluate these strategies to find the optimal one. Moreover, they can simultaneously minimise the energy cost of the extraction process by coordinating the industrial activity with the hourly grid energy cost fluctuation. Consequently, two artificial intelligence algorithms are proposed based on genetic algorithms and reinforcement learning. For the first, a population generator manages system’s restrictions based on real operative scenarios and then individuals change through time imitating natural selection, reproduction and mutation processes. For the second, a meta-heuristics policy is designed from state space and actions consisting on different heuristic strategies to explore potential solutions. Preliminary results evaluating the energy cost performance show that both algorithms reach similar solutions, registering the expected population features curve for the genetic algorithm but not illustrating a clear learning curve for the reinforcement learning study. This research is part of a LIFE-funded project (LIFE20 ENV/ES/000187). It was also supported by the Spanish Ministry of Science and Innovation (Project CEX2018-000794-S) and by AGAUR (projects 2017 SGR41 and 2021 SGR 00341). Finally, the first author gratefully acknowledges the Universitat Politècnica de Catalunya for the financial support of his predoctoral grant FPU-UPC, with the collaboration of Banco de Santander. Peer Reviewed Postprint (published version) 2025-11-21T06:31:49Z 2025-11-21T06:31:49Z 2024 Conference lecture http://hdl.handle.net/2117/446518 http://dspace.aeipro.com//handle/123456789/3625 http://creativecommons.org/licenses/by-nc-nd/4.0/ Open Access Attribution-NonCommercial-NoDerivatives 4.0 International Asociación Española de Ingeniería de Proyectos (AEIPRO)