Abstract: By optimizing the parameters of the importance sampling probability density function (IS-PDF) based on Kullback-Leibler (KL) distance, the cross-entropy based IS method (CE-IS) can significantly accelerate Monte Carlo simulation of power system reliability. However, KL distance is only one form of the f-divergence family which generalizes some other common distances besides the well-known KL distance. To explore the implementation of f-divergence in IS method, this paper proposed an optimal f-divergence IS method (OFD-IS). By minimizing the f-divergence between the ideal zero-variance probability density function (ZV-PDF) and IS-PDF, the uniform iterative updating formulas of IS-PDF parameters under the typical distance measure were derived. By using the uniform formulas in the variance minimization of the product of reliability index test function and likelihood ratio function, an iterative optimization method to determine the optimal distance in f-divergence family was proposed. The reliability evaluation of IEEE-RTS79 and IEEE-RTS96 system shows that the OFD-IS method can effectively improve the simulation efficiency, further explore the potential of efficiency improvement of IS method, and it is of great practical value for the rapid reliability evaluation of composite power systems.