Electronic Thesis and Dissertation Repository

Thesis Format

Monograph

Degree

Master of Engineering Science

Program

Electrical and Computer Engineering

Collaborative Specialization

Artificial Intelligence

Supervisor

Capretz, Miriam

Abstract

Data-driven predictive maintenance for heat, ventilation, and air conditioning (HVAC) systems has gained much popularity over recent years due to the increasing availability of integrated internet of things (IoT) sensors capable of reporting HVAC internal operational data. Most existing predictive maintenance methods are designed to analyse these internal operational data for maintenance decision making. However, these methods are not applicable to HVAC systems that are not equipped with internal IoT sensors. Consequently, we propose an AutoEncoder and Artificial Neural Network based HVAC Health Prognostics framework (AE-ANN-HP) that classifies the health condition of HVAC systems using only daily power consumption and outside temperature readings, both of which are easy to obtain for non-IoT enabled HVAC systems. AE-ANN-HP when evaluated with three types of autoencoders all show an increase in performance when compared to existing HVAC health prognostics methods in terms of classification accuracy.

Summary for Lay Audience

Heat, ventilation, and air conditioning (HVAC) systems account for a significant portion of energy consumption globally. Effective maintenance of HVAC systems can not only reduce the operational costs of the buildings but also improve human comfort of indoor environments. Data-driven predictive maintenance for HVAC systems has gained much popularity over recent years due to the increasing availability of integrated internet of things (IoT) sensors capable of reporting HVAC internal operational data. Most existing predictive maintenance methods are designed to analyse the internal operational data for maintenance decision making. However, these methods are not applicable to older or simpler HVAC systems that are not equipped with internal IoT sensors. Consequently, we propose an AutoEncoder and Artificial Neural Network based HVAC Health Prognostics framework (AE-ANN-HP) that classifies the health condition of HVAC system using only daily power consumption and outside temperature readings, both of which are easy to obtain for non-IoT enabled HVAC systems. The AE-ANN-HP framework is evaluated with three different types of autoencoders. Their performances are compared to existing HVAC health prognostics methods that are evaluated on the same dataset. The results show that all three AE-ANN-HP configurations yield higher classification accuracy than existing methods.

Creative Commons License

Creative Commons Attribution 4.0 License
This work is licensed under a Creative Commons Attribution 4.0 License.

Download

Share

COinS