The efficient and economical design and operation of individual small-scale distributed generation (DG) units has received considerable attention. It is now plausible to envision a future scenario where a large number of such units, spanning capacities from the kW to the MW scale, are deployed in a region. As distributed electricity generation using natural gas becomes more widespread, the dependency of the electric grid on the natural gas grid will increase. This concerns multiple interaction points along the two networks, and reaches beyond the natural gas demand of large-scale load following or baseload gas fired powerplants. More specifically, a better understanding is required of the potential changes in the dynamic behavior and interaction of the electricity and natural gas grids at or close to residential use sites (neighborhoods), where the small-scale DG infrastructure is likely to be located. In this paper, an optimization-based framework is developed for analyzing the operation of an ensemble of small-scale natural gas fueled DG units, and quantify their ability to flatten the electric grid load (i.e., reduce the peak demand) of the neighborhood that they serve. Our analysis relies on realistic energy use data and takes into account capacity limitations of the current natural gas distribution infrastructure, centralized vs. decentralized control of the DG unit operation, equipment durability considerations, heating preferences of home users, and seasonal effects. There is a substantial increase in natural gas consumption near consumers for all scenarios considered, which has implications for the control of the natural gas grid. We demonstrate the importance of having a centralized decision-making scheme when multiple distributed generation resources are present, and make recommendations for the optimal sizing of generators. (C) 2016 Elsevier Ltd. All rights reserved.