INVESTIGATION OF STRATEGIES FOR IMPROVING THE PERFORMANCE OF SMART THERMOSTAT-DRIVEN FDD METHODS

Abstract

This dissertation presents an investigation of two strategies for evaluating the performance of smart thermostat-driven fault detection and diagnosis (FDD) methods that are intended for vapor compression air conditioning (AC) systems in residential homes. Though there have been some research efforts within the last decade on smart thermostat-driven FDD, there are several challenges which have limited the advancement of the concept into a marketable technology. Therefore, the investigation in this dissertation is to explore some of the ways these challenges can be addressed.For the first strategy, a novel dynamic co-simulation model is developed using EnergyPlus and Modelica to gain a fundamental understanding of the coupling between building thermal response and AC system. The model is then validated and used to investigate the effectiveness of using smart thermostats for low-cost FDD in residential AC systems. Smart thermostats mea¬sures indoor air conditions which reflect the indoor air responses to both AC operations and other factors like weather and building gains. As these other factors are typically unmeasurable by smart thermostats and yet prone to variability, they pose potential disturbances which can affect AC operation. This makes estimating the impact of these disturbances (referred to as uncontrollable building load disturbances (UBLDs)) on AC operation in buildings and the performance of FDD methods critical. Therefore, in this study, it is essential to couple the dynamics of AC operation and building response in the development of the co-simulation model. Additionally, a novel automated calibration framework is also developed in this research to validate the coupled model. The validated model was then used to simulate seven UBLD cases and two prevalent faults with different severities. The study also proposed a set of metrics for evaluating the performance of FDD methods. With the simulation results and the proposed metrics, impact analysis on AC duty factor and enthalpy change were carried out. Results of the analyses showed that severe UBLDs can cause almost 9% increase in energy consumption. The results also showed that these UBLDs can cause false alarm rates in FDD up to 70% if appropriate thresholds and post-processing strategies are not used. Meanwhile, results on the impact of faults showed that with the simultaneous impact of UBLDs, only low charge faults up to 30% severity and low indoor airflow fault up to 60% severity can be confidently detected using duty factor as FDD feature. Meanwhile, the analyses and results also led to the proposal of another FDD feature (enthalpy change) which offers better sensitivity and a potential for diagnosis which the duty factor feature lacks. With this new feature, an automated FDD algorithm was developed, validated and successfully deployed in four test-homes where it was able to detect 30% undercharge fault and 30% low indoor airflow, as well as installation mismatch. Overall, this research work creates a new pathway for promoting smart thermostat FDD technology in the residential market within the US and beyond.