Using component surrogate models in the integrated design process for high-performance buildings

点击翻译按钮获取中文摘要

The buildings sector represents a significant portion of global energy consumption and greenhouse gas emissions. The design of buildings is a complex, dynamic, multi-disciplinary domain with long life-cycles and broad environmental impact. Simulation to support performance-based decision-making for building design and operation has evolved into an established practice, although traditional methods have struggled to adapt to the demands of net zero energy and emissions targets, especially in extending these often "premium services" to underserved and marginalized communities. Machine learning techniques can be used to train surrogate models that emulate discontinuous, multi-modal problem spaces with greater flexibility and computational efficiency than their high-fidelity counterparts. This research will apply Systems Decomposition, aided by qualitative study, to develop a conceptual framework of core functional requirements. Based on core research activities in the Energy in Cities lab at the University of Victoria, Surrogate models trained to fit components of the technical problem space will be aligned with the over-arching framework, and built up from high-fidelity simulations. This will create a powerful research platform suited for efficiently exploring some of the most pressing challenges that occur during early stage net zero building design, such as validation in the presence of significant uncertainties, advanced system configurations and complicated technical solutions, and demanding, non-traditional performance requirements capturing diverse community needs. Furthermore, through collaboration with the industry partner, tools and workflows will be developed that are grounded in practical application on real projects.