Developing and Calibrating a Cost-Effective Method to Determine Window to-Wall Ratio and Air-Conditioning Status of Existing Buildings (2022)
Authors: Abraham Philip, Bhavya Pathak, Shaily Gandhi, Rajan Rawal Abstract:
The study presents a cost-effective and scalable method to determine the Window to Wall Ratio (WWR) and A/C status of existing buildings from ground-view façade imagery. Object Detection Classifier deploying Faster Region-based Convolutional Neural Network (Faster R-CNN) is used to detect windows and buildings in visible images.
The detected elements are used by the second algorithm to calculate the Window to Wall Area Ratio (WWR) with a 40% variation from actual values. By superimposing the detected elements on corresponding thermal images of the building, a third algorithm is used to obtain the outside surface temperatures of windows and walls. Based on simulation study, a difference greater than 7°C between these values translates into air-conditioned zone. The research contributes to development of Urban Building Energy Model (UBEM) with LoD3. With such LoD, the UBEM model can provide quantitative and qualitative urban solutions for meeting the climate and energy goals.
Keyword: machine learning, urban building energy model, thermal imaging, air-conditioning, window wall ratio Presented at: The 5th International Conference on Building Energy And Environment (COBEE 2022), 25th to 29th July 2022, Concordia University, Montreal, Canada.
Supported by:
Ministry of New and Renewable Energy Govt. of India, Gujarat Energy Development Agency, U.S. Agency for International Development iNDEXTb (Industrial Extension Bureau) Govt. of Gujarat, Shakti Sustainable Energy Foundation, New Delhi
