Flexible Clean Propulsion Technologies

Predictive adaptive reactivity-controlled compression ignition for a dual-fuel marine engine: A model-in-the-loop study

Author

Xiaoguo Storm , Shamekhi Amir-Mohammad , Mohammad Raisi Esfarjani , Modabberian Amin , Vasudev Aneesh , Zenger Kai , Jari Hyvönen , Maciej Mikulski

Category

Publication channel

Keywords

Adaptive model predictive control, low-temperature combustion, RCCI, Real-time model

Year of the publication

2026

Citation

Storm, X., Shamekhi, A.-M., Raisi Esfarjani, M., Modabberian, A., Vasudev, A., Zenger, K., Hyvönen, J., & Mikulski, M. (2026). Predictive adaptive reactivity-controlled compression ignition for dual-fuel marine engines. Clean Propulsion Technologies Initiative. https://cleanpropulsion.org/wp-content/uploads/2026/05/Predictive-adaptive-reactivity-controlled-compression-ignition_120526.pdf

Language

English

Related to:

Abstract

Low-temperature, reactivity-controlled compression ignition (RCCI) combustion has proven instrumental in re- solving the long-standing trade-off between engine emissions and efficiency, particularly for heavy-dutyappli cations. However, RCCI has an inherent sensitivity to variations in-cylinder charge composition, such as fuel stratification, temperature gradients, and air-fuel mixing. This makes combustion behavior unpredictable and difficult to regulate using conventional control methods. This study presents an advanced multivariable model-based control design (MBCD) toolchain tailored for marine RCCI engines. Specifically, it introduces a real-time adaptive model predictive control (AMPC) strategy to regulate the indicated mean effective pressure (𝐼𝑀𝐸𝑃 ) and the crank angle at 50% mass fraction burned (𝐶𝐴50) by manipulating the total fuel energy and the blend ratio (𝐵𝑅) between the two fuels. The control framework is evaluated through model-in-the-loop (MiL) simulations with an experimentally validated high-fidelity UVATZ (University of Vaasa Advanced Thermo-Kinetic Multi-zone) model of a Wärtsilä 31DF engine combustor as the plant, and a physics-based linear real-time model (RTM) as an observer. The controller’s performance is benchmarked against a decentralized PI controller under various transient scenarios. Both controllers achieve comparable tracking of 𝐼𝑀𝐸𝑃 and 𝐶𝐴50, but the AMPC demonstrates faster 𝐼𝑀𝐸𝑃 response (within eight cycles), lower 𝐶𝐴50 steady-state error (maximum 0.45 crank- angle degree (CAD)), and reduced fuel consumption (2.7%). Additionally, AMPC’s receding-horizon framework and self-tuning features enhance robustness against unstructured uncertainties and parameter variations, marking a significant advancement over previously proposed predictive control strategies.