Partitioned model predictive control of LiU-Diesel engine air path based on recursive least squares

Modern diesel engine air-path systems operate the Variable Geometry Turbocharger (VGT) and Exhaust Gas Recirculation (EGR) to track set points for intake manifold pressure, as well as exhaust gas recirculation rate. This presents a challenging canonical case as it describes a highly non-linear multi-input multi-output (MIMO) system. This paper provides an examination of the development procedure for a segmented linear Model Predictive Controller (MPC) based on a data-driven approach. Leveraging the concept of linearization, we develop a series of computationally efficient prediction models, identified based on transient response data generated from the mean model LiU-Diesel. A series of local prediction models constitute a Linear Parameter-Varying (LPV) model covering the entire engine operating region. Subsequently, based on the LPV model, we design and validate a rate-based MPC. The results demonstrate that the MPC designed based on the data-driven LPV model is able to track specified set points and satisfy constraints on states and control.