In recent years, the impact of heavy rain disasters on the reliable power supply of distribution networks has become increasingly prominent, and it is extremely challenging to ensure the supply of important loads in the case of large-scale and multi-point serious faults. Rational deployment of emergency resources can help reduce the loss of important loads in the system, thereby improving the resilience of the distribution system. In this paper, a resilience planning method for emergency resources in the distribution system considering multiple uncertainties under heavy rain disasters is proposed, and a two-tier planning model for the resilience improvement of the distribution system is established considering the emergency power supply and emergency energy storage, with the upper model aiming to minimize the net cost of investment cost and load loss, and the lower model aiming to minimize the comprehensive load loss under multiple island scenarios. In addition, a stochastic simulation generation method for multi-island scenarios in distribution network under heavy rain disasters is proposed, and based on the probability and severity of each scenario, a scene reduction method based on clustering algorithm is proposed to serve the scene screening of the lower-level optimization model. Finally, through the analysis of the actual 62-node system of China's distribution network, it is shown that the method proposed in this paper can effectively improve the supply capacity of the distribution network for important loads while taking into account the economy.