A Methodology for Automating the Implementation of Advanced Control Algorithms Such As Model Predictive Control on Large Scale Building HVAC Systems
Loading...
Date
2018-01-23
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
Building operations consume about 40% of the total energy consumption in the US, with Heating Ventilation and Air-Conditioning (HVAC) comprising a significant portion of it. HVAC system of a typical commercial building consists of several components – cooling towers, chillers, Air Handling Units (AHUs), fans dampers, etc. Improving the performance of these components has the potential for large energy and cost savings. Implementing better control methodologies for regulating these components can be the first step in this direction for building managers, as it requires minimal retrofitting. Advanced control methodologies such as Model Predictive Control (MPC) can help realize this potential.
Three reasons for the inefficient operation of traditional control methodologies are identified in this dissertation, the improper tuning of PI controllers for nonlinear systems, a decentralized control architecture that doesn’t perform any global optimization, and lack of planning for future operating conditions. The dissertation makes a contribution towards addressing the aforementioned areas of inefficiencies by offering a solution in the form of alternative control architectures such as cascaded control and optimal control algorithms such as MPC.
The dissertation first provides the results of a survey that demonstrates the widespread nature of the phenomenon of hunting (undesired oscillations) in building HVAC systems. An algorithm to detect the presence of hunting in real time is proposed and implemented on data obtained from real buildings on the campus of Texas A&M University. A description of the cascaded control architecture is provided along with a simulation example to show how it can mitigate the problem of hunting.
The dissertation addresses the other two reasons for inefficient operation, namely the decentralized control architecture and lack of planning for the future by proposing a method that would allow the process of implementing advanced control algorithms such as MPC to be automated and easy to implement. Currently implementing MPC on large scale building HVAC systems remains a major bottleneck as it requires the development of models that are accurate and easy to compute. This dissertation makes a contribution towards this front by proposing a modeling algorithm that can be automated and scaled to systems comprising hundreds of components. The modeling algorithm is verified using data from a real office building. Static models are developed for the Air Handling Unit (AHU) pressure subsystem which includes the AHU fan and Variable Air Volume (VAV) boxes serving conditioned air to the rooms. In addition to the static models, dynamic models are developed for the AHU temperature subsystem comprising a heat exchanger that uses chilled water (CHW) to cool the air passing through the AHU. Dynamic models are also developed for the temperatures of 9 of the 11 rooms of the office building, thereby demonstrating that the proposed algorithm can be implemented for multi-zone systems.
The dissertation also makes a contribution towards the implementation of advanced controls by providing a method by which the black-box models can be used to implement MPC on building HVAC systems. MPC using models developed from the proposed modeling algorithm is applied to a high-fidelity simulation model of the office building. Results of the simulation show that MPC can provide significant energy savings over the traditional control algorithms.
Description
Keywords
System Identification, Automated Modeling, Model Predictive Control, Large Scale Building HVAC Systems