Smart Grids, the next-generation power grids, are  equipped with modern technologies to address the increasing demand on the quality of the power supply in the traditional power grids. Power grid is a critical infrastructure with reliability being one of its key quality attributes. Current reliability analysis methods for power grids focus on the physical level of the grid and do not consider multiple layers of the smart grids, such as software components and services. In our research, we aim at developing an approach for reliability analysis that would take into account multi-layered nature of smart grid by incorporating the physical components, software components and service usage scenarios. In this talk we introduce an initial formal model based on Stochastic Reward Nets that allows integrated reliability modelling of multiple smart grid layers.