Metaheuristic optimization of electro-hybrid powertrains using machine learning techniques
- Christopher Bacher, MASTER’S THESIS, Verfassung der Arbeit
Parameter optimization for Hybrid Electric Vehicles (HEVs) is very time consuming due to the necessity to evaluate a simulation model. This bottleneck is usually removed by using metamodels as surrogates for the simulation. We consider metamodels for longitudinal dynamics simulation, which simulate a vehicle following a given driving cycle. Typical “top-down” metamodels are parametrized by both the HEV model and the driving cycle. We propose a novel bottom-up metamodelling scheme only parametrized by the HEV model and discuss preliminary results.