AI-Assisted approach for building energy and carbon footprint modeling

• Main Authors: Chih-Yen Chen, Kok Keong Chai, Ethan Lau
• Format: Article
• Language: English
• Published: Elsevier 2021-09-01
• Series: Energy and AI
• Subjects: Building energy simulation and benchmarking; Carbon footprint; Occupant density; Artificial intelligence; Long short-term memory (LSTM); Smart city
• Online Access: http://www.sciencedirect.com/science/article/pii/S2666546821000458

This paper proposes an energy and carbon footprint modelling using artificial intelligence technique to assess the impact of occupant density for various types of office building. We use EnergyPlus to simulate energy consumption, and then estimate the related CO₂ emissions based on three years (2016–2018) of Actual Meteorological Year (AMY) weather data. Various occupant densities were used to evaluate the annual energy consumption and CO₂ emission. In this work, a robust deep learning technique of long short-term memory (LSTM) model was established to predict the time-series energy consumption and CO₂ emissions. A power exponential curve was suggested to correlate the behaviour of annual energy and CO₂ emission for occupant densities range from 10 to 100 m²/person for each office building type. The results of LSTM model show high prediction performance and small variations within the three types of office building data, which can be applied to the similar building model to predict and optimise energy consumption and CO₂ emission.