Model öngörülü kontrol için hızlı hesaplama stratejisi

This thesis focuses on slow operation issues of the existing Model Predictive Control (MPC) algorithms and proposes a method to speed the slow response issues (rise time, settling time and prediction) of them up due to the computation load of online optimization without reducing the computation load. In the proposed method of this thesis, the original MPC strategy is warped using an integer time scaling factor in time domain. The aim is to speed this strategy up by exploiting the speeds of fast microprocessors. For this aim an integer time scaling factor is selected according to the limit of the used microprocessor. The solution of the plant is contracted and sampled by this scaling factor. Standard MPC problem is based on this modification of the plant. First elements of fast computed control law vectors are dilated until original online optimization sampling period is reached again and applied to plant at each sampling period. Modified MPC strategy, which is called Upsampled MPC (UMPC), settles faster to desired output than original MPC strategy and also satisfies faster prediction at each sampling period. Since the method is more compatible with fast microprocessor and speed of the UMPC results from microprocessor, slow operation issue of MPC algorithms is solved for different time scaling factors by this method. Therefore, control energy of UMPC is lower. There also exists free times between online optimization sampling periods of original MPC and UMPC strategies. These free times can be used for improvement of existing MPC algorithms.