Robust constrained optimization for RCCI engines using nested penalized particle swarm

Abstract

Reactivity controlled compression ignition (RCCI) is a promising combustion concept which uses two fuels to combine high thermal efficiencies and low engine-out NOx and soot emissions. The combustion concept relies on controlled auto-ignition and is sensitive for changing injection pressure, fuel quality, etc. Consequently, modeling and control of this complex combustion concept is not straightforward. In this work, Gaussian process regression is used to arrive at a data-driven model for a gasoline-diesel RCCI engine. This data-driven model is employed in a robust optimization approach that uses a nested particle swarm optimization. The designed (feedforward) control inputs maximize the efficiency of the RCCI engine while satisfying safety and emissions constraints under various disturbed conditions. In the simulation study, robust performance is obtained, and the robust efficiency is very similar to the efficiency under nominal condition.