Energy demand is rapidly growing, resulting in rising energy cost, depletion of fossil fuels, energy shortage, environmental concerns and network overloading. All these problems need new and novel ways for redressal. Conventional grids may not address these problems. However, smart energy distribution grids offer solutions such as distributed generators (DGs), demand response (DR) and energy hub (EH) approach. This work focuses on the application of the DGs and the DR equipped EH paradigm, for efficient energy management, in residential and commercial buildings. The EH control functions in two layers. Primary stratum termed as micro EH control resides in customer premises to attain building level stakes such as energy consumption cost and greenhouse gas (GHG) emission minimization. However, utility’s benefit termed as network load deviation, remains unattended. Wide range of load variations may cause network instability and quality issues. Secondary layer termed as utility owned macro energy hub (MEH) control minimizes network load deviation; however, building incentives forfeit. In other words, customer-utility benefits are non-linearly related and cannot be attained simultaneously. Therefore, a strategy has been required to resolve this non-linearity. This thesis proposes bi-level MEH controls for residential and commercial buildings, to simultaneously attain bilateral stakes. Moreover, customer-utility models are developed and analyzed for both sectors. As real-life energy systems function under uncertain conditions, therefore, proposed micro as well as macro EH control modules have also been validated under random solar irradiance as well as unplanned electric and natural gas (NG) network outages. Under these uncertainties, two types of bi-level MEH controls are formulated: 1) risk neutral and 2) risk averse. Former control disregards negative effects of uncertainties; however, later reduces the adverse effects of uncertainties on the performance and the resilience, by incorporating conditional value at risk (CVaR) in the objective function. Proposed linear constrained optimization models have been mapped in flower pollination algorithm (FPA) as it shows superior performance in literature. Obtained results have been compared to the ones presented in literature and found meaningful improvement.