Emission-constrained LPV-MPC control for marine RCCI engines based on hybrid grey-box modeling

Abstract

As global decarbonization policies tighten, accurate prediction and control of combustion and emission behavior in Reactivity Controlled Compression Ignition (RCCI) engines are crucial to achieving high efficiency and ultra-low emissions in the maritime industry. A hybrid grey-box framework combining a physics-based combustion model with LSTM networks is constructed to accurately predict CA50, IMEP, MPRR, and emissions including CO and THC. Additionally, an LPV state-space model is then identified by using LS-SVM with fuel quantity as scheduling parameters for fast and adaptive prediction under dynamic conditions. Furthermore, a multi-objective MIMO MPC controller is designed to track CA50 and IMEP, while constraining MPRR, COVIMEP, and emissions. Moreover, GA and PSO methods are employed to optimize MPC weights and reference trajectories, thereby enhancing control performance. The proposed control strategy can reduce CA50 and IMEP tracking errors by 35.6 % and 29.1 %, respectively, and reduce CO and THC emissions by up to 12.7 %, compared to a conventional PI controller, while ensuring smooth actuator and constraint satisfaction. This framework could provide an effective approach for intelligent RCCI engine management, thus supporting cleaner combustion and enhanced efficiency in the maritime sector.