A Graph Propagation-Based Automatic Differentiation Method for Power Electronic and Electric Machine Systems Simulation

This paper proposes an automatic differential method to enable the effective Taylor-series based flexible integration algorithm for power electronics and electric machine systems simulation. The proposed approach decomposes the system equations into linear and non-linear parts and encapsulates the non-linear part through controlled-source interfaces. Furthermore, this paper introduces a graph propagation-based automatic differentiation method (GP-AD), which enables the automatic generation of higher-order derivatives for nonlinear portion. To evaluate the efficiency of the proposed approach, a case study on a power traction system of a high-speed train is presented. The results indicate that the proposed algorithm achieves a 6-fold speedup compared to the commercial software, demonstrating the effectiveness of the proposed method.