Optimizing Model Base Predictive Control for Combustion Boiler Process at High Model Uncertainty

Abstract

This paper proposes a multi-objective evolutionary algorithm for optimizing model base predictive control (MBPC) tuning parameters applied to the boiling process. The multi-objective evolutionary algorithms are able to incorporate many objective functions that can simultaneously meet robust stability and performance that can satisfy control design objective functions. These promising techniques are successfully implemented to stabilise MBPC at the implications of different levels of model uncertainties. The Pareto optimum technique is able to overcome the problem of trapping the standard genetic algorithms (SGAs) in the local optimum when using the LQ as the objective functions at the price of high model uncertainty. Introducing robust stability and performance objective functions has successfully improved the search procedure for MBPC tuning variables at high model uncertainty.

Keywords

• model base predictive control
• multi-objective evolutionary algorithm
• control tuning