There have been significant advances in communication technologies over the last decade, such as cellular networks, Wi-Fi, and optical communication. Not only does the technology impact peoples’ everyday lives, but it also helps cities prepare for power outages by collecting and exchanging data that facilitates real-time status monitoring of transmission and distribution lines. Smart grids, contrary to the traditional utility grids, allow bi-directional flow of electricity and information, such as grid status and customer requirements, among different parties in the grid. Thus, smart grids reduce the power losses and increase the efficiency of electricity generation and distribution, as they allow for the exchange of information between subsystems. However, smart grids is not resilient under extreme conditions, particularly when the utility grid is unavailable. With the increasing penetration of the renewable energy sources (RES) in smart grids, the uncertainty of the generated power from the distributed generators (DGs) has brought new challenges to smart grids in general and smart microgrids in particular. The rapid change of the weather conditions can directly affect the amount of the generated power from RES such as wind turbine and solar panels, and thus degrading the reliability and resiliency of the smart microgrids. Therefore, new strategies and technologies to improve power reliability,sustainability, and resiliency have emerged. To this end, in this thesis, we propose a novel framework to improve the smart microgrids reliability and resiliency under severe conditions. We study the transition to the grid-connected operational mode in smart microgrids,in the absence of the utility grid, as an example of emergency case that requires fast and accurate response. We perform a comparative study to accurately predict upcoming grid-connected events using machine learning techniques. We show that decision tree models achieve the best average prediction performance. The packets that carry the occurrence time of the next grid-connected transition are considered urgent packets. Hence, we per-form an extensive study of a smart data aggregation approach that considers the priority of the data. The received smart microgrids data is clustered based on the delay-sensitivity into three groups using k-means algorithm. Our delay-aware technique successfully reduces the queuing delay by 93% for the packets of delay-sensitive (urgent) messages and the Packet Loss Rate (PLR) by 7% when compared to the benchmark where no aggregation mechanism exists prior to the small-cell base stations. As a mitigation action of the utility grid unavailability, we use the electrical vehicles (EVs) batteries as mobile storage units to cover smart microgrids power needs until the utility grid recovery. We formulate a Mixed Integer Linear Programming (MILP) model to find the best set of electrical vehicles with the objective of minimum cost. The EVs participating in the emergency power supply process are selected based on the distance and throughput performance between the base station and the EVs