The full text of this item is not available at this time because the student has placed this item under an embargo for a period of time. The Libraries are not authorized to provide a copy of this work during the embargo period, even for Texas A&M users with NetID.
An Investigation of the Energy, Cost, and Comfort of Adaptive Model-Predictive Controls for Residential HVAC Systems
Abstract
The building sector constitutes the largest portion of energy consumption worldwide, surpassing the industrial and transportation sectors. In the United States, buildings alone account for approximately 40% of total energy (with residences contributing 22% and commercial buildings consuming 18%) and are responsible for 40% of carbon emissions. As part of the 2030 U.S. action plan to combat climate change, a target of 50-52% reduction in greenhouse gas (GHG) emissions has been set, compared to the 2005 level. Therefore, quite clearly, reducing energy consumption in buildings, especially in the residential sector, holds significant potential to contribute towards achieving these ambitious energy and emission mitigation goals.
Among the energy consumption in U.S. buildings, the Heating, Ventilation, and Air-Conditioning (HVAC) system, which provides a desirable environment for occupants living inside, is responsible for the majority of the total energy used. Compared to building automation systems that are extensively utilized in commercial buildings, HVAC control systems integrated for current residential buildings are limited. The central forced-air system is widely used in regular homes throughout North America, where manual, programmable, and connected ones (commonly known as smart thermostats) are the three main categories of thermostats. The first two categories are found in traditional thermostats, whereas connected thermostats offer additional capabilities such as remote setpoint operation via smartphones. However, occupants often struggle to determine the appropriate setpoint that will ensure their comfort while achieving energy savings, primarily because they are unable to consider the adaptation to several uncertain scenarios. One such scenario is the degradation of cooling/heating capacity caused by air conditioner (AC) unit faults. This degradation can hinder the HVAC system’s ability to maintain desired thermal requirement, especially when the cooling/heating load is high. Additionally, the emergence of the smart grid, coupled with increased on-site renewable generation, offers the opportunity to leverage new grid features, such as electricity price variations, to optimize energy consumption and reduce residents' energy bills. This ability to respond is highly appealing.
Therefore, this research develops a cost-effective HVAC control framework for smart homes that has the capability of connecting fault detection and diagnosis and corresponding fault-adaptive control to provide an acceptable indoor thermal environment with lower energy consumption. Additionally, the framework is designed to be responsive to grid signals, aligning with grid requirements. To ensure its applicability in regular homes, a practical and user-friendly approach is generated, making it easy to implement and utilize.
Fault modeling of residential HVAC systems has been established, which encompasses the creation of a fault library and the construction of modeling methods. This fault library comprises 24 common faults categorized into five types, commonly found in residential buildings. This workflow is demonstrated by injecting multiple faults into EnergyPlus models of prototype buildings. Fault modeling results are beneficial to understand the impact of faults in residential HVAC systems, serving as a foundation for generating fault detection and fault-adaptive control algorithms.
Fault Detection and Diagnosis (FDD) algorithms are then generated for residential air-source Vapor Compression Cycle (VCC) systems, identified as the most impactful fault based on previous fault modeling investigations. Seven potential faults along the refrigerant line are considered. A quantitative comparison analysis is conducted for three rule-based FDD algorithms, utilizing open-source laboratory data and exclusive field data. Based on the accuracy rates observed for various faults, a hybrid FDD approach is recommended, with temperature measurements being the key determinant.
When faults are detected that affect the HVAC capacity, strategies for pre-cooling or preheating homes are introduced. Model Predictive Control (MPC), recognized for its efficacy in slow and non-linear dynamic systems such as HVAC systems, stands out as a control methodology. This method capitalizes on the thermal mass of buildings, enabling precooling to offset the consequences of capacity degradation or to shift the building load. Simplified operational rules are then extracted from the offline implementation of a large-scale MPC system using a rule extraction method. Artificial Neural Network (ANN) is used to develop models for MPC, while optimization is achieved using Differential Evolution (DE). Rule extraction, facilitated by a Decision Tree (DT), emulates the MPC results, enabling the optimized control trajectory to be implemented to a certain extent in regular homes. These MPC-informed rules are computationally efficient and can be executed online in residential HVAC controllers. Using the same computer setup, the average runtime for MPC is 870 minutes over a ten-day period, while the MPC-informed Rule-Based Control (RBC) finishes in an average of 0.2 minute. This makes the MPC-informed RBC significantly faster than the original MPC.
Finally, field tests are conducted to verify the effectiveness and efficiency of the developed FDD and adaptive control strategies. Two identical lab homes are utilized for this purpose. Lab Home A functions as the control home, while Lab Home B is the baseline home. For the FDD strategy demonstration, the heat pump system's FDD accuracy achieves 90.18%, counting all cases (including fault-free ones), with no false alarms. The detection accuracy stands at 96.4%. In terms of improving thermal comfort, unmet hours could be reduced from 2.05 ℉-hour (Lab Home A) to 0.22 ℉-hour (Lab Home B), thereby enhancing residents' thermal comfort under HVAC system malfunction or during heatwave conditions.
Subject
Residential buildingFault modeling
Fault detection and diagnosis
Model predictive control
Rule extraction
Lab testing
Citation
Yang, Tao (2023). An Investigation of the Energy, Cost, and Comfort of Adaptive Model-Predictive Controls for Residential HVAC Systems. Doctoral dissertation, Texas A&M University. Available electronically from https : / /hdl .handle .net /1969 .1 /200068.