Multiplexed Real-Time Optimization of HVAC Systems
複雜空調系統的多路控制策略
Student thesis: Doctoral Thesis
Author(s)
Related Research Unit(s)
Detail(s)
Awarding Institution | |
---|---|
Supervisors/Advisors |
|
Award date | 3 Aug 2017 |
Link(s)
Permanent Link | https://scholars.cityu.edu.hk/en/theses/theses(7aef072b-ebdb-4abf-907f-5da38937d003).html |
---|---|
Other link(s) | Links |
Abstract
The real-time, optimal control of heating, ventilation, and air-conditioning (HVAC) systems is an efficient tool for effective operation and for improving energy efficiency. The real-time, optimal control of HVAC systems needs to regularly optimize the decision variables or set-points of the local control loops, taking into account the interactions between HVAC components with the help of system models. After optimization, the set-points are updated as soon as possible. In an HVAC system, the set-points for several local control loops have a significant influence on the overall energy performance of the system. Therefore, the real-time optimization (RtOpt) of those set-points has been widely studied. However, due to the non-linear dynamics of the HVAC system as well as the constraints associated with the system’s operation, RtOpt always suffers from a heavy online computational load when those set-points are optimized simultaneously. To overcome this problem, multiplexed real-time optimization (MRtOpt) has been developed. Unlike conventional RtOpt, which optimizes all set-points simultaneously, MRtOpt optimizes and updates only one set-point at a time, but with a higher optimization frequency.
The first major drawback of the MRtOpt strategy is that it requires frequent resetting of the set-points; this introduces artificial disturbances within the control loop, as well as disturbances in other local control loops that interact with this loop, which may, in turn, significantly affect the stability of the whole system. Conventional reset methods follow a step reset or a rate-limited change. They present similar problems because they lack a systematic study. Secondly, the real-time optimal control of HVAC systems relies on the system’s performance models to predict its performance. Several performance models of HVAC systems are already developed that rely on modeling techniques, such as physical or white box, black box, and grey box modeling. A conventional RtOpt strategy generally incorporates the system’s white box models are subject to several simplification assumptions. These performance models are not accurate, and since they need a full mathematical description of the process, they are not suitable for an RtOpt strategy because they tend to increase the computational load and memory demand.
It is important to select an accurate, reliable, and computationally efficient performance model of an HVAC system. Therefore, for this study, the performance model of HVAC systems was selected and developed with respect to these criteria in order to ensure reliability and effectiveness of the developed MRtOpt strategy for HVAC systems. To identify a better way to reset the set-points (other than a step reset or a rate-limited reset), it is important to study the transient behaviours of the affected local control loops that are triggered by resetting the set-points. Therefore, disturbance models are developed with a subspace identification method (SIM) using a canonical variant-analysis (CVA) approach to study the transient behaviour of HVAC systems due to the resetting of set-points. The SIM is a black box system-identification technique that has a tendency to model single-input/multi-output (SIMO) systems without structural parameterization. These disturbance models are used in the synthesis of a new, model-based optimal-reset method because of their intelligibility and intrinsic capability. This new method is called the degree-of-freedom- (DOF) based set-point reset method, and it provides freedom in terms of the magnitude of smaller resets to approach the desired set-point with minimum excitation of the disturbance model. To illustrate the performance, a dynamic-simulation test bed of a virtual HVAC system was synthesised using the MATLAB-TRNSYS simulation software. It was demonstrated that in comparison with the conventional step reset and rate-limited reset methods, a DOF-based set point reset can significantly improve on the transient performance of an HVAC system’s local control loops. Therefore, for the stability enhancement of MRtOpt, it was integrated with a DOF-based set-point reset method to update the set-points instead of the conventional step reset or rate-limited reset. The control performance of the integrated strategy was investigated using case studies. The results demonstrated that approximately 10% of the energy saving was achieved by the proposed method, compared to a method without RtOpt. When compared with the conventional RtOpt method, the proposed method resulted in a computational load reduction of approximately 70%, and a reduction of over 26% in the tracking errors of the local control loops.
To summarise this study, it can be stated that a new RtOpt mechanism was developed for HVAC systems. This involved the development of a reliable, dynamic-simulation test bed of an HVAC system; the selection and development of a performance model of the system for performance prediction; identification of disturbance models to describe the transient behaviour of local control loops activated by set-point reset; the development of a new, model-based reset method, which is a DOF-based method, for the set-point resetting; and finally the implementation of this DOF-based set-point reset method with an MRtOpt strategy. It was demonstrated that the proposed MRtOpt with the DOF-based set-point reset can improve system stability and provide efficient good energy performance. Also, it is more suitable for practical application because of the reduction in the online computational load.
The first major drawback of the MRtOpt strategy is that it requires frequent resetting of the set-points; this introduces artificial disturbances within the control loop, as well as disturbances in other local control loops that interact with this loop, which may, in turn, significantly affect the stability of the whole system. Conventional reset methods follow a step reset or a rate-limited change. They present similar problems because they lack a systematic study. Secondly, the real-time optimal control of HVAC systems relies on the system’s performance models to predict its performance. Several performance models of HVAC systems are already developed that rely on modeling techniques, such as physical or white box, black box, and grey box modeling. A conventional RtOpt strategy generally incorporates the system’s white box models are subject to several simplification assumptions. These performance models are not accurate, and since they need a full mathematical description of the process, they are not suitable for an RtOpt strategy because they tend to increase the computational load and memory demand.
It is important to select an accurate, reliable, and computationally efficient performance model of an HVAC system. Therefore, for this study, the performance model of HVAC systems was selected and developed with respect to these criteria in order to ensure reliability and effectiveness of the developed MRtOpt strategy for HVAC systems. To identify a better way to reset the set-points (other than a step reset or a rate-limited reset), it is important to study the transient behaviours of the affected local control loops that are triggered by resetting the set-points. Therefore, disturbance models are developed with a subspace identification method (SIM) using a canonical variant-analysis (CVA) approach to study the transient behaviour of HVAC systems due to the resetting of set-points. The SIM is a black box system-identification technique that has a tendency to model single-input/multi-output (SIMO) systems without structural parameterization. These disturbance models are used in the synthesis of a new, model-based optimal-reset method because of their intelligibility and intrinsic capability. This new method is called the degree-of-freedom- (DOF) based set-point reset method, and it provides freedom in terms of the magnitude of smaller resets to approach the desired set-point with minimum excitation of the disturbance model. To illustrate the performance, a dynamic-simulation test bed of a virtual HVAC system was synthesised using the MATLAB-TRNSYS simulation software. It was demonstrated that in comparison with the conventional step reset and rate-limited reset methods, a DOF-based set point reset can significantly improve on the transient performance of an HVAC system’s local control loops. Therefore, for the stability enhancement of MRtOpt, it was integrated with a DOF-based set-point reset method to update the set-points instead of the conventional step reset or rate-limited reset. The control performance of the integrated strategy was investigated using case studies. The results demonstrated that approximately 10% of the energy saving was achieved by the proposed method, compared to a method without RtOpt. When compared with the conventional RtOpt method, the proposed method resulted in a computational load reduction of approximately 70%, and a reduction of over 26% in the tracking errors of the local control loops.
To summarise this study, it can be stated that a new RtOpt mechanism was developed for HVAC systems. This involved the development of a reliable, dynamic-simulation test bed of an HVAC system; the selection and development of a performance model of the system for performance prediction; identification of disturbance models to describe the transient behaviour of local control loops activated by set-point reset; the development of a new, model-based reset method, which is a DOF-based method, for the set-point resetting; and finally the implementation of this DOF-based set-point reset method with an MRtOpt strategy. It was demonstrated that the proposed MRtOpt with the DOF-based set-point reset can improve system stability and provide efficient good energy performance. Also, it is more suitable for practical application because of the reduction in the online computational load.
- building energy efficiency, real-time optimization, degree-of-freedom- (DOF) based set-point reset, heating, ventilation, and air conditioning (HVAC), control stability