Toward the application of a machine learning framework for building life cycle energy assessment
Thesis
Overview
Research
Identity
Other
View All
Overview
abstract
Nearly half of the global annual energy supply is consumed by buildings in their construction, operation, and maintenance, indicating an enormous potential to minimize the carbon footprint. During its life cycle, a building consumes energy in the form of embodied and operational energy. Embodied energy (EE) is expended in processes during construction, this includes extraction of raw material, transportation, manufacturing, etc. Operating energy (OE) is spent on operating and maintaining the building to ensure occupant comfort. Studies show that improving the operational efficiency of a building may have serious implications for EE. Building life cycle energy assessments (LCEA) is, therefore, essential to understanding the dichotomy between EE and OE. Traditionally, data-driven approaches such as simulation-based optimization techniques are used for design space exploration. Literature shows that these data-driven approaches are error-prone, time-consuming, and computationally expensive, and fail to provide real-time feedback to the user. Besides, EE and OE assessment tools are disjointed and suffer from the issues of interoperability. These limitations restrict design space exploration, which eventually hinders the design decision-making process. In recent years, increased availability and accessibility of large-scale data have made machine learning (ML) techniques a popular choice for building performance assessment. In this context, numerous articles have developed prediction models to assess or optimize OE. While this work is significant, there remains a lack of studies that have utilized ML techniques for building LCEA mainly due to the lack of a large-scale LCEA database. This study proposed to generate a simulation-based building energy dataset for different building typologies using a parametric framework. The synthetically generated database was then used in the development of the ML model. The artificial neural network (ANN) model developed in this research would provide quick and reliable results related to the buildings EE intensity and OE intensity. Furthermore, the application of the ANN prediction model was demonstrated using a case study. The experimental results of the case study show that the developed prediction model achieved high prediction performance using minimal inputs that are available during the early design phase. The results of this research indicate that ML techniques can indeed be used to instantaneously estimate building LCE performance. The practical implementation of this research would help designers with no experience in using simulation tools select design options with minimal LCE consumption.
