摘要：AbstractHybrid electric drivetrains are an example of systems that can enable discrete switching between states with very distinct continuous behavior. For the design of these drivetrains, the required computation time for optimizing the control over a driving cycle is critical. To reduce computation time, we propose two methods to improve the implementation of dynamic programming by reducing the number of grid points that computationally demanding sub-models are evaluated for. The proposed methods do not require surrogate models and can be applied to arbitrary drivetrain topologies. A case study on a parallel, a series-parallel, and a power-split hybrid drivetrain is performed, and a reduction in computation time of up to 66% is shown.
