Data-driven model predictive control of buildings

Weng, Kui 2020. Data-driven model predictive control of buildings. PhD Thesis, Cardiff University. Item availability restricted.

Preview	PDF - Accepted Post-Print Version Download (5MB) | Preview
	PDF (Cardiff University Electronic Publication Form.) - Supplemental Material Restricted to Repository staff only Download (107kB)

Abstract

Buildings account for 30% of the final global energy and 28% of the total carbon emissions in the world. Heating, ventilation and air-conditioning (HVAC) systems can consume up to 60% of the total energy consumption in buildings. Improving the energy efficiency of HVAC systems is important in reducing carbon emissions and mitigating risks associated with global climate change such as overheating of indoor environments. Another benefit of improving the energy efficiency of HVAC systems is to save energy cost for building owners. Many previous studies focused on the design and retrofit for improving building energy efficiency, but few of them looked into how to improve the building operation. As the primary building energy system, Commercial HVAC systems are complex because of the interaction of a large number of sub-systems and uncertainties resulting from the interactions of building mass, thermal inertia, weather and occupancy. The application of Model Predictive Control (MPC) has received significant attention in the last few years from researchers and the industry to the control and management of building energy systems. Despite increasing research on using MPC for improving the energy efficiency of HVAC systems, few of them utilise flexibilities such as time of use (ToU) and killowatt Max (kWmax) control. This research investigates how the control of building elements (such as windows) and HVAC systems could improve energy efficiency and thermal comfort. This study has been divided into two parts based on three case studies. The first part of the study demonstrates a physics-based case study that assesses the impact of climates on the indoor environment and how the control of window openings for natural ventilation can reduce overheating risk in current and future climates. The results find bedrooms are easier to suffer overheating risks than living rooms but increasing openings for natural ventilation is more effective in reducing overheating hours in bedrooms. By opening 20% of window area for natural ventilation, the results show that 2%, 17% and 45% of the total 108 dwellings’ bedrooms are overheated in the 2030s, 2050s and 2080s, compared to living rooms with 30%, 60% and 89%. In the 2030s, increasing the window opening area ratio from 20% to 80% can reduce the number of dwellings with overheating risk in bedrooms from 32 to 14, but find nearly no change in living rooms. However, the passive control of building elements such as windows, blinds and overhangs has limitations in adapting dwellings to climate change. With a maximum window area for opening plus blinds and 2-meter overhangs, it can still not eliminate overheating risk in most UK cities in the 2080s. After demonstrating the limitations of the control of building elements in future climates, the second part of the study introduces two case studies which turn to study the iv optimisation of controls for HVAC systems in a residential and a commercial building. The research goes towards the development of data-driven MPC controllers for the two buildings. A sensor network has been established for building energy metering and environmental monitoring in the residential building to enable remote control of the heating system with the MPC controller. It is found that the MPC controller can improve thermal comfort by allowing more hours with room temperatures within the design comfort band. In the commercial case study building, a data-driven MPC controller has been developed, running optimal control of 9 indoor units per 15 minutes to maintain indoor temperatures within the design comfort band. It proposes a demand response method to minimize energy cost by integrating with ToU and kWmax use cases. The study finds that MPC could take advantage of energy tariffs and flexibility by shifting the loads from high-demand periods to low-demand periods. With the data-driven MPC, it could reduce the peak energy consumption by up to 36% and the peak power by about 15%.

Item Type:	Thesis (PhD)
Date Type:	Completion
Status:	Unpublished
Schools:	Engineering
Uncontrolled Keywords:	Data-driven model; Model predictive control; Machine learning; HVAC system; Near real-time control; Overheating risk.
Date of First Compliant Deposit:	5 March 2021
Last Modified:	04 Mar 2022 02:30
URI:	https://orca.cardiff.ac.uk/id/eprint/139294

Actions (repository staff only)

Edit Item