Online adaptive and intelligent control strategies for multizone VAV systems

Nearly one half of the total energy used in buildings is consumed by HVAC systems. With escalating cost of energy, several energy efficiency strategies have been implemented in buildings. Among these, the use of VAV systems, and improved method of controlling such systems have received greater attention. This thesis is devoted to design and development of online adaptive control strategies which will be augmented with optimal and intelligent-control algorithms. The considered VAV system consists of zone air temperature control, discharge air temperature control, water temperature control and air pressure control loops. Online adaptive control strategies are developed for these control loops. In order to design reliable online controls a robust RLS identification algorithm for estimating the parameters of the modeled processes is developed. It is shown that this algorithm avoids wrong estimation and requires fewer variables compared with classical RLS techniques. Three different online control strategies were designed. These are: a robust optimal control algorithm (ROCA), a simplified optimal adaptive control (SOAC) for FOPDT systems, and a two-loop adaptive control strategy which improves both temperature and airflow regulations in VAV systems. ROCA is an on-line optimal proportional-integral plus feedforward controller tuning algorithm for SISO thermal processes in HVAC systems. It was optimized by combining the H {592} based PI tuning It is shown that the two-loop adaptive control strategy has both stronger robustness to time-varying thermal loads and lower sensitivity to airflow rate changes into other zones. The developed control strategies were tested by simulation and experiments in a VAV laboratory test facility which uses existing energy management control systems used in commercial buildings. Also, an adaptive neural network controller is developed. The proposed controller was constructed by augmenting the PID control structure with a neural network control algorithm and an adaptive balance parameter. Simulation results show that the proposed controller has stronger robustness, improved regulation and tracking functions for FOPDT type plants compared to classical PID controllers. Experiments were conducted to verify the characteristics of the developed controller on the DAS in a two-zone VAV test facility. Applications of the developed control strategies to different control loops in VAV system were demonstrated by conducting several experimental tests under realistic operating condition