Simulation based on standard models is often used as part of the engineering design process to test theories and exercise new concepts before actually placing them into operation. In order to tackle the problem of likely widespread catastrophic failures of electric power grids, an autonomously reconfigurable power system will have to rely on wide-area communication systems, networked sensors, and restorative strategies for monitoring and control. Standard practice is to use simulation of a small number of certain historical test systems to test the efficacy of any proposed design. We believe this practice has shortcomings when examining new communication system ideas. In this paper we develop the means for producing power grids with scalable size and randomly generated topologies. These ensembles of networks can then be used as a statistical tool to study the scale of communication needs and the performance of the combined electric power control and communication networks. The topological and system features of the randomly generated power grids are compared with those of standard power system test models as a "sanity check" on the method.