The distributed, user-specific, bidirectional communication promised by smart grids allows for novel improvements to the current power grid's naive power allocation and pricing strategies. We present a modeling framework for a smart microgrid power system which employs a bidding mechanism to distribute power efficiently and intelligently to residential houses, which are in turn modeled as smart homes with smart appliances. We model the power demand within these smart homes using queuing theory. We provide an algorithm that guarantees the houses' powers (which are a function of the houses' bids), to converge to the unique game-theoretic optimal power allocation (i.e. Nash equilibrium). We present numerical simulations in several different regimes to highlight the convergence properties.