Study of Adaptive Neural Network Model-based Predictive Control for HVAC Systems
DescriptionEnergy consumption of buildings accounts for a significant portion of the energy “demography” of developed global cities (e.g. Hong Kong, Beijing, Shanghai, Tokyo, New York and London, etc). For example, commercial building energy consumption alone in Hong Kong in 2013 was around 284 PJ/annum, contributing 2 tonnes of CO2per person. Heating, Ventilation and Air-conditioning (HVAC) systems consume at least 75% of the total building energy consumption in HK. It has been shown that inefficient control of HVAC systems account for half of all inefficiencies in the system. Hence, there is still a considerable margin for improvement through optimal control of HVAC systems.This research aims to develop optimal control techniques to synthesize an intelligent and costeffective solution for building energy usage without compromising individual level of comfort. The key objective is to device an optimal supervisory control solution for HVAC systems to minimize energy use and improve efficiency of the overall system.Usually, HVAC systems respond slowly to demand. Therefore Model Predictive Control (MPC) that is well suited for slow response systems will be employed for decision making. In view of the fact that the behaviour of HVAC systems is highly nonlinear and complex, an adaptive neural network (aNN) model based on PENN (Probabilistic Entropy based Neural Network) will be synthesized to capture its dynamic behaviour. The aNN model will predict the system response and update the MPC. In addition, delay (dead time) compensator for local loop control will be designed for effective system operation. The aNN model based predictive control (aNNMPC) which works on the principle of receding horizon will be developed using MATLAB. This will also deal directly with constraints and eventually it can replace the current HVAC control schemes. A simulation test bed of an HVAC system will be built in the TRNSYS simulation environment and the Matlab aNNMPC will be used as the control for the system. This will enable an assessment of the effectiveness of aNNMPC and comparisons with conventional control schemes.It is an anticipated that the new control scheme will enhance robustness, accuracy, reliability and efficiency as well as improve system operation and reduce maintenance costs.
|Effective start/end date||1/01/19 → …|