Learning-Based NMPC for Hydrogen–Diesel Engines

Overview

This project focuses on the development of machine learning–enhanced nonlinear model predictive control (NMPC) for hydrogen–diesel dual-fuel engines, with the goal of improving combustion stability, reducing emissions, and maximizing hydrogen utilization under real-time constraints.

Motivation

Hydrogen-assisted combustion introduces strong nonlinearities, cycle-to-cycle variability, and safety-critical constraints that challenge conventional control approaches. This project addresses these challenges by integrating data-driven prediction models within a constrained NMPC framework.

Methodology

LSTM-based time-series models for engine output prediction
Control-oriented feature selection and preprocessing
NMPC formulation with safety, stability, and actuator constraints
Real-time optimization and deployment considerations

Experimental Validation

Multi-cylinder engine test platform
Cycle-to-cycle closed-loop implementation
Evaluation metrics: tracking error, combustion stability, constraint violations

Outputs

Journal and conference publications
MATLAB / Python control and modeling pipeline
Experimental datasets (where applicable)

Author et al., “Title of the paper,” Journal / Conference, Year.