Machine Learning Integrated with Model Predictive Control for Imitative Optimal Control of Compression Ignition Engines

The high thermal efciency and reliability of the compression-ignition engine makes it first choice for many applications. For this to continue, a reduction pollutant emissions is needed. One solution use Machine Learning (ML) Model Predictive Control (MPC) minimize fuel consumption, without adding substantial computational cost controller. ML developed in paper both modeling performance imitating behaviour Linear Parameter Varying (LPV) MPC. Using support vector machine-based linear parameter varying model emissions, predictive controller implemented 4.5 L Cummins diesel engine. This online optimized MPC offers advantages minimizing NOx consumption compared baseline feedforward production To reduce MPC, deep learning scheme designed mimic behavior reducing at constant load by imitative similar that however, requires 50 times less computation time when optimization.