Search results

The purpose of this research is to design a robust high-performance nonlinear multi-input multi-output heating, ventilation and air conditioning (HVAC) system controller for temperature and relative humidity regulation. Buildings are complex systems which are subjected to many unknown disturbances. Further complicating the control problem is the fact that, in practice, buildings and their systems have static nonlinearities such as power saturation that make stability difficult to guarantee. Therefore, in order to overcome these issues, a control system must be designed to be robust (performance insensitive) against uncertainties, static nonlinearities and effectively respond to unknown heat load and moisture disturbances.

A state of the art nonlinear inverse dynamics (NID) technique is combined with a genetic algorithm (GA) optimisation scheme in order to improve robustness against uncertainty in the system's modelling assumptions. The parameter uncertainty problem is addressed by optimising the control system parameters over a specified range of uncertainty. The NID control structure provides further robustness with effective disturbance handling and a stability criteria that holds in the presence of actuator saturation.

The proposed method delivers significantly more energy efficient performance whilst achieving improved thermal comfort when compared with a current industry standard HVAC controller design such as proportional-integral-derivative. The expected excellent response to disturbances is also demonstrated.

This method can easily be extended to account for other parameters with a specified uncertainty range.

This research presents a method of optimised NID controller design which can be easily implemented in real HVAC controllers of building energy management systems with a high degree of confidence to provide high levels of thermal comfort whilst significantly reducing energy usage.

A novel HVAC optimised NID control strategy using the robust inverse dynamics estimation feedback control topology with GA optimisation for improved robustness and tuning over a range of parameter uncertainty is described, designed and its performance benefits shown through simulation studies.