Tuning DMC controller using multi-objective optimization for the CIC2018 Benchmark Challenge

In this paper, a multivariable Dynamic Matrix Control (DMC) tuning using multi-objective optimization (MO) is presented and evaluated on the one-staged refrigeration cycle model described in the CIC2018 benchmark challenge – an adaptation of the benchmark proposed in the PID18 Conference. The MO approach takes advantage of the relative indexes given by the benchmark, allowing knowledge about how much the DMC parameters influence control performance. Results of the MO approach are shown using Level Diagrams (LD), a tool for visualization and analysis of multidimensional Pareto fronts. Consequently, different DMC tunings are selected for different trade-offs between the relative indexes. Selection procedure shows how relevant an analysis of the Pareto Front can be in the decision-making stage for multivariable controller tuning.