This study presents a comparison of different approaches for the simulation of HEV fuel consumption. For this purpose a detailed 1D-CFD model within an HEV drivetrain is compared to a ‘traditional’ map-based combustion engine model as well as different types of simplified engine models which are able to reduce computing time significantly while keeping the model accuracy at a high level.
First, a simplified air path model (fast running model) is coupled with a quasi dimensional, predictive combustion model. In a further step of reducing the computation time, an alternative way of modeling the in cylinder processes was evaluated, by replacing the combustion model with a mean value model. For this approach, the most important influencing factors of the 1D-CFD air path model (temperature, pressure, A/F-ratio) are used as input values into neural nets, while the corresponding outputs are in turn used as feedback for the air path model. However, while the computing speed of the simulation can be further increased, this model type loses its predictiveness, compared to detailed combustion models.
The performance of said engine models is evaluated within a HEV drivetrain model. Results for the New European Driving Cycle as well as the Artemis Urban Driving Cycle are shown.