RUB » Bibliotheksportal

Kasimir Forth

• Whole building energy simulations are an established method for predicting the operational energy performance of new constructions. Nevertheless, these tools are mostly expert systems and setting up a Building Energy Model (BEM) still requires a lot of knowledge and effort. Often, current approaches, which transform Building Information Models (BIM) models for energy simulation in early design stages, don’t take model-specific information related to energy properties into account or need to be enriched manually in the simulation tools. Sometimes, only generic construction sets or program types are assigned to the model, so the simulation results don’t correspond with the specific information of the architectural BIM model. In this paper, we are proposing a novel methodology which creates BEM based on BIM by automatically enriching HVAC (Heating, Ventilation, Air Conditioning) program types for each room using Natural Language Processing (NLP). Based on the architectural room names as input, we propose to use multilingual large language models (LLM) to identify the semantically most similar program type to automatically enrich the model with the relevant information. Finally, we are testing the proposed approach using an office building as a case study.