Adaptive Horizon Adjustment for Model-Predictive Control

Internal Model type controllers such as Dynamic Matrix Control and Model Algorithmic Control attempt to minimize the ill effects of control effort constraints by feedback around the process model. The controls are chosen to minimize an integral of the "distance" between the desired and achievable model response. An alternate method of selecting the controls is to require that the state of the model match the state of the trajectory generator at selected times in the future. The adaptive horizon algorithm presented in this work selects controls so that the model state and the trajectory generator state match after n+1 time periods, where n is the order of the minimum phase portion of the process model. To accommodate constraints the algorithm adaptively adjusts the size of the time periods so that none of the controls beyond, the initial step violate constraints. The algorithm has been applied to a number of single-input, single-output systems and shows good tracking of the desired trajectory with only a modest computational effort.