Weather events can cause power outages and damage the grid infrastructure. Resilience is the ability of the power grid to quickly recover from disruptions such as weather events. Enhancing the resilience of the distribution systems is required for continued grid operation. A solution to enhance the resilience of a distribution system is to proactively deploy resources (e.g., backup generators, repair crews, additional conductors) in the most vulnerable regions to provide faster service restoration. In this paper, first, we propose a method to model the impact of extreme weather events on power grids. Then, we use Q-learning to identify the worst impact zones while considering possible propagating paths of extreme weather. Finally, we develop a game-theoretic approach to allocate resources to the worst impact zones at the minimum cost. Simulations are conducted on a modified IEEE 123-bus test system. The results prove the effectiveness of the proposed work on enhancing grid resilience.