Defining generation parameters with an adaptable data-driven approach to construct typical meteorological year weather files

Anahita Bigtashi,Ashleigh Papakyriakou,Bruno Lee

DOI: https://doi.org/10.1016/j.enbuild.2023.113781

IF: 7.201

2024-01-01

Energy and Buildings

Abstract:Weather files play a critical role in building energy simulation by providing crucial information on the environmental condition. Typical meteorological year (TMY) weather files are often used in building energy simulation to approximate the long-term performance. These files are generated using long-term weather data and are meant to reflect the typical weather conditions within the dataset. Common generation methods used to construct TMY weather files typically rely on expert-defined universal generation parameters. However, the weather parameters and weighting factors used to generate these files, termed generation parameters, often overlook variations in climate and applications. This study introduces a data-driven approach to define the generation parameters used in the Sandia method to produce TMY weather files. The study aims to enhance the generation method by providing a framework to define the generation parameters which considers data availability, climate and application. The proposed methodology employs machine learning and feature importance to define the generation parameters used to construct TMY weather files. The generated TMY weather files are evaluated for a Montreal case study using building energy demand. The results show an improvement in the representativeness of the long-term average building energy demand for the weather file generated using the model-defined generation parameters.

energy & fuels,construction & building technology,engineering, civil