TOM.D: Taking advantage of microclimate data for urban building energy modeling

Abstract

Urban Building Energy Modeling (UBEM) provides a framework for decarbonization decision-making on an urban scale. However, existing UBEM systems routinely neglect microclimate effects on building energy consumption, potentially leading to major sources of error. In this work, we attempt to address these sources of error by proposing the large scale collection of remote sensing and climate modeling data to improve the capabilities of existing systems. We explore situations when remote sensing might be most valuable, particularly when high quality weather station data might not be available. We show that lack of access to weather station data is unlikely to be driving existing errors in energy models, as most buildings are likely to be close enough to collect high quality data. We also highlight the significance of Landsat8’s thermal instrumentation to capture pertinent temperatures for the buildings through feature importance visualizations. Our analysis then characterizes the seasonal benefits of microclimate data for energy prediction. Landsat8 is found to provide a potential benefit of an 8% reduction in electricity prediction error in the spring and summertime of New York City. In contrast, NOAA RTMA may provide a benefit of a 2.5% reduction in natural gas prediction error in the winter and spring. Finally, we explore the potential of remote sensing to enhance the quality of energy predictions at a neighborhood level. We show that benefits for individual buildings translates to the regional level, as we can achieve improved predictions for groups of buildings.

Keywords

• Urban microclimate
• Remote sensing
• Machine learning
• Land surface temperature
• Climate model
• Urban heat island