AI-based energy modelling and forecast for heating, ventilation and air conditioning (HVAC) systems

• Main Author: Lim, Celia Kar Cheng
• Other Authors: Chau Yuen
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Engineering
• Online Access: https://hdl.handle.net/10356/181620
• Institution: Nanyang Technological University

The building sector is a significant energy consumer globally, with Heating, Ventilation, and Air Conditioning (HVAC) systems accounting for 40-60% of energy consumption in commercial buildings. Despite efforts under Singapore's Green Plan 2030 to reduce energy use, the country's commercial building energy consumption has grown by 25% from 2008 to 2019 and is projected to rise further. Traditional HVAC systems face challenges in incorporating external factors like weather and occupancy patterns, which affect energy efficiency. This project aims to address these limitations by deploying artificial intelligence (AI) based techniques, specifically the Long Short-Term Memory (LSTM) deep learning method, to generate a surrogate model for energy modelling, hourly consumption forecasting as well as predicting other performance metrics such as systems’ peak demand, and zonal temperatures and humidity in HVAC systems. Firstly, the surrogate model was trained on data outputs generated from a Building Energy Modelling (BEM) and Simulation software, Openstudio. The building and HVAC system data used for the simulation was from an actual manufacturing plant that is located in Tuas, Singapore and consisted of a complex HVAC system. Hence, the data used for the simulation represents a real-world scenario. Finally, the trained LSTM model is able to achieve an R² value of >90% on AHU consumption and room parameters, explaining their variance at >90% all the time. This AI-based approach offers potential improvements in HVAC system efficiency and energy savings.