Bp 342 en
Fast Numerical Methods for Model Predictive Control
Author: Pavel Otta
Predictive control (MPC) is modern method for controlling industrial processes. Natural advantage of this method according to its principle is possibility of considering limits of controlled system and then for defined criteria optimal control action can be found. Thus, main element of control is the optimalization process, where mostly quadratic program is being solved (when qaudratic criteria of optimality which may be stated as a energetic criteria is selected). MPC leads to optimalization problem with constraints, where no analytic result exists as for some optimalizations without constraints. Short sampling interval systems control has to be solved as fast as possible by appropriate algorithms. However, for simple constraints f.e. the box shape, effective algorithms with ones aid of controlling even fast sampled systems is possible exist. In terms of this thesis new controller with this algorithm was implemented.