Automatic Evaluation and Optimization of Generic Hybrid Vehicle Topologies using Dynamic Programming

Abstract

The optimal design of a hybrid drive train is a challenging problem. The design space of different topology configurations is large, and evaluating the fuel efficiency of a concept requires a control strategy. Effective methods to readily compare multiple designs are still unavailable. We propose a framework to automatically evaluate and optimize hybrid electrical vehicle topologies. This is a first, crucial step for the exploration of the full design space. The optimal controls are computed using Dynamic Programming (DP). DP is often deemed too slow for practical use, but we suggest some improvements to reduce the computational complexity significantly. A second contribution is the automatic generation of a causal model from a topology description, built from a component library. By using a causal model, we avoid solving the model equations implicitly, further reducing the computational load. Using a parallel hybrid topology as an example, we validate the methodology and show that the proposed method is suitable for property optimization.