Development and testing of Model Predictive Controllers for an automotive organic Rankine cycle unit

Two-thirds of the energy produced by an internal combustion engine (ICE) is lost into waste heat through the coolant and the exhaust gas; hence, studying Waste Heat Recovery (WHR) systems is of vital importance. The organic Rankine cycle (ORC) is a powerful system to recover low-grade heat and transform it into electrical energy. This thesis aimed at developing and testing a Model Predictive Control (MPC) system that ensures a safe operation of a system that constitutes an ICE bottomed by an ORC unit. The experimentation was carried out at the DTU Mekanik laboratories and was divided into different campaigns. Firstly, to study the plant behavior, steady-state and dynamic characterizations were accomplished. The latter was useful to obtain transfer function models for the MPCs at different vehicle speeds. Secondly, Proportional-Integral (PI) controllers and MPCs qualities were evaluated thanks to three performance indices while the engine was following a testing cycle. The MPC model was derived at 90km/h. Afterward, a test campaign aimed at optimizing the tuning parameters of the MPC cost function and at evaluating their influence on the plant response. Finally, the controllers that performed best were tested on a World harmonized Light-duty vehicles Testing Cycle (WLTC) to characterize their operation under realistic driving conditions. The results showed that MPCs were more suitable for the task than PIs due to their better ability to operate the plant in safe conditions, and to their best performance indices when subjected to the testing cycles as well as to the WLTC. Nevertheless, MPCs have to be further optimized to follow the homologation cycle. Future experimentations could be based on be exploiting multi-model systems constituted of two or more MPCs or obtaining the MPC model from other working points.