Multi-objective PID Controller Tuning for Multi-model Control of Nonlinear Systems

Abstract

Multi-model approaches for non-linear control usually consist in partitioning the operating space of a given non-linear process, in such a way, it is possible to approximate a local linear model for each partition. That allows using standard control and tuning techniques for each local linear model. After that, a global controller merging the set of local controllers is defined to control the original non-linear process. This global controller merges the local controllers, given that a single controller will exhibit different trade-offs in its performance in different operating regions; therefore, a single controller could exhibit a deficient performance in the overall operating space. In this paper, we exploit this fact from a multi-objective optimisation point of view, to tune a single controller using a parametric multi-model, looking for a suitable controller with a preferable trade-off along with the overall operating space.