Given the often-limited availability of real electricity network data, this paper presents methodology for the synthesis of multilevel electricity networks for use in applied network failure and risk analysis. The proposed algorithm is capable of producing networks that preserve a number of important spatial and topological properties of real-world networks including the multilevel structure of subsystems, the geographic distribution of network nodes, the node degree distribution, and the networks spatial connectivity. The algorithm is capable of integrating both synthetic and real data from a range of sources to produce spatially and topologically continuous representations. The flexibility of the algorithm is demonstrated through the synthesis of a regional-scale electricity network. The practicality of the algorithm, in terms of providing new data to conduct applied risk and resilience studies of interdependent infrastructures, is demonstrated through the synthesis of a unique representation of the national integrated electricity network for England and Wales, bridging the transmission, subtransmission, and distribution scales and consisting of more than 160,000 nodes.