Efficient control and Fault Detection and Isolation in Building HVAC systems

Abstract

My thesis work is focused on advancing the state of the art in smart building monitoring and control problems. A major part of my thesis will develop combined model- and data-driven models for the energy systems of buildings and apply them for energy monitoring and optimization, fault diagnosis, and fault-adaptive control. In past work, I have already developed initial machine learning-based models for energy optimization. In this paper, I am focusing on developing a framework using bond graphs to build a model of the HVAC system and understand how different factors affect the measurement variables of this system. This information will help us do subsequent fault diagnosis by studying the characteristics of the system behavior represented by the temporal causal graph. It will also provide the basis for fault-adaptive control of buildings.