A hybrid switching predictive controller based on bi-level kernel-based ELM and online trajectory builder for automotive coldstart emissions reduction

In the current study, a novel intelligent control algorithm is proposed for the efficient reduction of coldstart hydrocarbon (HC) emissions of an automotive engine. The proposed intelligent controller inherits the computational advantages of two advanced techniques, namely hybrid switching predictive controller and extreme learning machine (ELM). The proposed model-based predictive controller has a switching behavior which enables it to be used for a deliberate control of the considered engine's behavior during the coldstart period. As the hybrid controller requires a hybrid model for processing, a bi-level ELM is used which is composed of two independent ELMs for the calculation of exhaust gas temperature (Texh) and engine-out HCs. A novel online trajectory builder is also introduced to make sure that the control commands are adjusted according to the actual behavior of the engine. Through simulations, the computational potentials of ELM for designing hybrid switching controllers are demonstrated. Furthermore, it is indicated that the proposed hybrid switching controller can afford qualified results for the considered case study. To verify the authenticity of the obtained results, two well-known controllers, namely nonlinear model predictive controller (NMPC) and an optimal controller based on the Pontryagin's minimum principle (PMP), are taken into account. The numerical experiments also host an additional step which evaluates the impact of prediction horizon (HP) on the performance of the controller. Based on all of the mentioned simulations, the efficacy of hybrid switching predictive controller with bi-level ELM (HSPC-BELM) is verified.