PREDICTIVE CONTROL OF NONLINEAR DYNAMIC PROCESSES

Predictive control can be applied if the reference value of the proces s is known in advance and the deterministic disturbances can be predic ted. A cost function defined in the future horizon is minimized. The c ontrol signal is calculated for a control horizon, but only the first one is applied and the procedure is repeated (receding horizon strateg y). Processes with mild analytical nonlinear characteristics are consi dered. The possible process models are either nonparametric (linear, H ammerstein, and Volterra weighting function series) or parametric ones (generalized Hammerstein, parametric Volterra, and bilinear models). The algorithms of the optimal and suboptimal predictive control based on the nonparametric and the parametric models mentioned are derived. Several simulations present how effective these methods are. The adapt ive case is dealt with as well.