Model-based intelligence multi-objective globally optimization for HCCI engines

摘要

Modern engines feature a considerable number of adjustable control parameters. With this increasing number of Degrees of Freedom (DoF) for engines, and the consequent considerable calibration effort required to optimize engine performance, traditional manual engine calibration or optimization methods are reaching their limits. An automated engine optimization approach is desired. In this paper, a self-learning evolutionary algorithm based multi-objective globally optimization approach for a Homogenous Charge Compression Ignition (HCCI) engine is developed. The performance of the HCCI engine optimizer is demonstrated by the co-simulation between an HCCI engine Simulink model and an Strength Pareto Evolutionary Algorithm 2 (SPEA2) based multi-objective optimizer developed in Java. The HCCI engine model is developed by integrating the physical gas exchange model, in-cylinder volume model and statistical combustion model. The model has been validated from 1500 rpm to 2250 rpm with different Indicated Mean Effective Pressure (IMEP). The model is able to simulate the performance of in-cylinder pressure, Indicated Specific Fuel Consumption (ISFC) and Indicated Specific Hydrocarbon (ISHC) emissions with acceptable accuracy in real-time within a wide engine operation window. The SPEA2 optimizer has been validated by the classic evaluation function SRN with constrains. The validation results show that the optimizer can find the Pareto Front of SRN efficiently. The introduced Intelligence optimization is an approach to optimize the engine ISFC and ISHC simultaneously by adjusting the engine actuators’ settings automatically through SPEA2. For this study, the HCCI engine actuators’ settings are Intake Valves Opening (IVO), Exhaust Valves Closing (EVC) and relative air to fuel ratio (λ). The co-simulation study and experimental validation results show that the intelligent multi-objective optimizer can find the optimal HCCI engine actuators’ settings with acceptable accuracy, and much lower time consumption than usual.