Abstract: In order to improve the initialization stability and convergence speed of large-scale AC / DC hybrid power grid, the artificial neural network is used to fit its complex impedance, harmonic, and control characteristics, and then nested into the three-phase fundamental power flow calculation framework based on the improved node method for iterative correction. At the same time, in order to improve the initialization efficiency of large-scale AC / DC hybrid power grid, a parallel initialization calculation method based on multi-port equivalence is proposed. The simulation results of IEEE33 bus and 330 bus AC / DC hybrid power grid with photovoltaic power stations are used to verify the effectiveness of the proposed method.