查看文章

In this paper a general procedure for tuning multivariable model predictive controllers (MPC) with constraints is presented. It has been applied to tune the control system of an activated sludge process control in a wastewater treatment plant. Control system parameters are obtained by solving a mixed sensitivity optimization problem, defined in terms of the H_∞ norms of different weighted closed loop transfer functions matrices of the system, and a set of constraints, some of them expressed using the l₁ norm. The use of multiple linearized models for the control allows for the specification of many robust performance criteria. The mathematical optimization for tuning all controller parameters is tackled in two iterative steps due to the existence of integer and real numbers. First, integer parameters are obtained using a special type of random search, and secondly a sequential programming method is used to tune the real …