Research On The Method Of Predictive Control Based On The Multiscale Theory

Model predictive control is a model-based advanced control technology that has been already applied widely in industrial process industries. Classical control theory and modern control theory are based on exact mathematical modeling of the object to be controlled. In practice, however, the object is usually a complex process which is multi-variable,high-order and time varying. So it is difficult to establish its exact mathematic model. Model predictive control does not demand a high accurate model; hence it can be applied to complex plants rather efficiently. In recent years, with the development of many kinds of the control methodologies, such as intelligent control, PID control, robust control and hybrid systems control, the theory of model predictive control has also obtained rapid developments. In practical applications, for large-scale complex industry systems, traditional approaches in model predictive control (MPC) suffer from one serious disadvantage: the need for infinite horizon for robust stability and performance which will bring in high computational complexity and large computational burden. Aiming at solving the above problem, our work will develop as follow:1) Based on the wavelet transformation of time-domain models, the multiscale models are developed on dyadic whose nodes are used to index the values of any variable and establish multi-scale dynamic systems.2) A new multiscale parallel MPC algorithm is proposed based on the multiscale model, the new algorithm possesses some good characteristics: not only reduces the complexity and increases running speed of algorithm, but also ensures robust stability. Finally, the difference in performance between traditional model predictive control (MPC) and multiscale model predictive control (MSMPC) is illustrated by simulation results.3) The controllability and observability of multiscale domain was defined in multiscale systems. In this paper, the methods of Gelam matrix criterion and Order criterion will be used to testify controllability and observability of multiscale systems.