Generation of synthetic models of gas distribution networks with spatial and multi-level features

This paper addresses the generation of models of distribution gas networks with realistic topological, spatial and technical features. The proposed algorithm offers a novel methodology for reproducing networks with multiple pressure levels. The procedure is tailor-designed for gas distribution networks, although minor ad-hoc modifications could extend its application to virtually any other physical network infrastructure. A probabilistic approach is followed, in which Gaussian Mixture Models (GMM) are used for spatial placement of the synthetic nodes and distance-based criteria are formulated to establish connections among them. Different pressure levelsare separately generated and connected in a hierarchical fashion via pressure reduction stations, while independent subnetwork islands are identified by introducing a novel clustering scheme. The strengths and feasibility of the approach are verified on a real test case via comparison of specific structural properties and results highlight a fine agreement between the topology and the geographical distribution of the real and synthetic networks. It follows that the algorithm offers a first effective solution toward the generation of spatially-embedded network models to treat gas networks and structures with multiple hierarchical levels.