Creating Realistic Synthetic Power Distribution Networks based on Interdependent Road Infrastructure Rounak Meyur 1* , Madhav Marathe 2 , Anil Vullikanti 2 , Henning Mortveit 2 , Virgilio Centeno 1 and Arun Phadke 1 1 Department of Electrical and Computer Engineering, Virginia Tech 2 Biocomplexity Institute & Department of Computer Science, University of Virginia {rounakm8, virgilio, aphadke}@vt.edu, {marathe, vsakumar, henning.mortveit}@virginia.edu Abstract Physical inter-dependencies between networked civil infrastructures such as transportation and power system network are well known. In order to analyze complex non-linear co-relations between such networks, datasets pertaining to such real in- frastructures are required. Such data are not read- ily available due to their sensitive nature. This work proposes a methodology to generate realis- tic synthetic distribution network for a given ge- ographical region. The generated network is not the actual distribution system but is very similar to the real distribution network. The synthetic net- work connects high voltage substations to individ- ual residential consumers through primary and sec- ondary distribution networks. The distribution net- work is generated by solving an optimization prob- lem which minimizes the overall length of network and is subject to the usual structural and power flow constraints. The work also incorporates identifica- tion of long high voltage feeders originating from substations and connecting remotely situated cus- tomers in rural geographical locations. The pro- posed methodology is applied to create synthetic distribution networks in Montgomery county of south-west Virginia, USA. The created networks are validated for their structural feasibility and abil- ity to operate within acceptable voltage limits under average load demand scenario. 1 Introduction In the present day world, human behavior, social networks and civil infrastructures are closely intertwined to each other. Recent works such as Zeng et al. [2019], Duan et al. [2019] show how decisions undertaken by human behavior impact the load on infrastructure which in turn affects the following human decisions. Recent advances in artificial intelligence and computational science has opened new opportunities to analyze complex non-linear network dynamics in interdepen- dent networks such as Adiga et al. [2020]. * rounakm8@vt.edu In order to study coupled infrastructures, a central chal- lenge is the lack of realistic data sets. Such datasets should have detailed representations of each network as well as the interactions across infrastructures and the interactions with the human population. Over the past few years, Gegner et al. [2016], Birchfield et al. [2017], Atat et al. [2019] have tried to address the problem of generating realistic power networks so that they are openly available to researchers for studying complex phenomena. However, these works are limited to power transmission network where the impact of individual consumer decision is not prominent. Problem This works proposes a methodology to generate re- alistic synthetic distribution networks for a geographical loca- tion using open source information. We further aim to gener- ate a network which follows structural (radial configuration) and operational constraints (voltage and power flows within limits) of a real distribution system. 1.1 Related Works Several methodologies have been proposed to generate ran- dom distribution network topologies based on statistical dis- tributions. A graph theory based approach to construct a synthetic distribution network is studied in Schweitzer et al. [2017] using statistical distributions of graph attributes (node degree, hops from source etc.). In Atat et al. [2019] a stochas- tic geometry based approach is proposed to place transform- ers in synthetic power networks. However, these works do not consider the various power system operation constraints such as node voltage and edge flows. Furthermore, the gen- erated networks are not optimal in terms of network loss and cost of installation which is always the primary consideration of a distribution company. Kadavil et al. [2016] proposed a bottom-up approach is used to build a synthetic distribution network from a sub- station and thereafter populating it with consumer loads. A mixed integer second order conic program (MI-SOCP) based optimization problem is formulated by Trpovski et al. [2018] to generate a synthetic medium voltage network for a geo- graphical region. However, the location of loads are consid- ered to be the zip code centers and loads are aggregated for each zip code which results in an aggregated synthetic dis- tribution network of the region. The generated synthetic net- works do not replicate the distribution of actual consumers in a geographical region. Thus, similar networks would be gen- arXiv:2001.09130v1 [eess.SY] 24 Jan 2020