Study On Generalized Predictive Control And Their Simulations

Generalized predictive control (GPC) generated in the 1980s is a new computer control algorithm, which is a representative predictive control algorithm. Generalized predictive control preserves the predictive control's advantages of model predictive, receding option and feedback correction, and has good control precision and robustness. So, GPC has gained considerable attention since the generation of GPC, and has been widely applied to industrial processes.GPC is studied here in the base of other's study and the thesis is organized as follows: Firstly, constrained GPC algorithm is studied. Nonlinear programming can be converted into linear programming by Goal programming, and which is solved by basic line algorithm in the thesis. Goal programming algorithm as a multi-objective optimization strategy can distinguish soft and hard restriction directly, and its solution is not depending on the chosen initial conditions.Secondly to solve the difficulty of identifying MIMO systems, adopt subspace system identification algorithms to identify state space model of system. The GPC base on state space model is deduced by myself, so the GPC and subspace system identification algorithms can combine better.Finally, study a GPC algorithm for a kind of usual nonlinear system. This kind of nonlinear system can be converted into linear parameter-varying system and identified parameter-varying on line, then GPC can be used directly. Compared with other nonlinear GPC algorithm, this algorithm preserves the same control effect, increases the count and advances the track speed when the working point changing.