Long Range Predictive Optimal Control Law With Guaranteed Stability for Process Control Applications

Abstract

The Generalized Predictive Control law has been successfully applied in industrial applications but has limitations on nonminimum phase processes for some of the most obvious choices of cost-functions. The Weighted Predictive Control law proposed here is new and avoids these difficulties while maintaining a similar philosophy. It ensures guaranteed stability when the control weighting tends to zero, even if the system is nonminimum phase. The solution is relatively straightforward and is very suitable for process control applications. The cost-function includes dynamic weighting terms on both output and control signals so that robustness properties can be frequency shaped.