BIM-driven artificial intelligence solutions for energy efficiency : from theoretical framework development to photovoltaic energy integrated planning

• Ano de defesa: 2025
• Autor(a) principal: ALVES, Josivan Leite
• Orientador(a): PALHA, Rachel Perez
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: eng
• Instituição de defesa: Universidade Federal de Pernambuco
• Programa de Pós-Graduação: Programa de Pos Graduacao em Engenharia Civil
• País: Brasil
• Palavras-chave em Português: Building Information Modeling; Artificial Intelligence; Solar Energy; Sustainability; Times series
• Link de acesso: https://repositorio.ufpe.br/handle/123456789/65735

The Architecture, Engineering, Construction, and Operations (AECO) sector has benefited from data generation and management through Building Information Modeling (BIM). This increased availability of data can be fundamental for driving innovation when analyzed with Artificial Intelligence (AI) models. In this context, the pursuit of smart sustainable projects has become a strategy to promote sustainable development and reduce dependence on non- renewable resources. The relevance of this topic is evident in the fact that buildings consume approximately 40% of global energy and are responsible for 33% of greenhouse gas emissions, largely due to low energy efficiency. Given this problem, this research aims to develop automated BIM-driven AI solutions to support the energy planning of sustainable buildings, focusing on the simulation and optimization of photovoltaic systems in both new and retrofit projects. The investigation adopts a multi-method approach, combining a systematic literature review, the development of deep learning algorithms, and the implementation of automated BIM modeling processes. First, a mapping of the integration between BIM and AI was performed, identifying application domains, problems addressed, results achieved, and fundamental capabilities of both technologies. Next, a BIM-driven deep learning algorithm was proposed and tested to estimate photovoltaic energy production, relating solar irradiation time series with data automatically extracted from BIM models. The results demonstrate that the proposed approach (EnergyBIM.AI) enables the automatic quantification of solar energy production and the potential reduction of CO2 emissions. Furthermore, the research proposes a process to support the selection and placement of solar panels in BIM models. Thus, this thesis demonstrates that the integration of BIM and AI can transform energy planning in the AECO sector, offering automated processes capable of increasing energy efficiency, reducing emissions, and supporting the design and retrofit of sustainable buildings. The main contributions of this thesis are: (i) offering a framework for integrating BIM and AI applied to smart projects; (ii) proposing an approach for energy forecasting based on deep learning and automation in BIM; and (iii) provide evidence to support the design and retrofit of sustainable buildings.