The availability of innovative tools for fast computation of power flow (PF) on extended grids assumes a fundamental role in the current context of the electricity system. With the increasing size and complexity of power systems and their frequent stressful operating conditions, conventional PF solvers may be unable to provide an adequate and fast response.

A possible promising method is to build network models using graph databases and to apply graph computing techniques for the development of faster and more efficient algorithms. In this work, real distribution networks have been modeled using the graph database TigerGraph from their semantic description in the IEC CIM standard.

Under this representation, power flow equations have been solved using a graph algorithm similar to Google PageRank, for which node voltages are computed sending parameters through messages between neighbouring nodes. Compared to MATPOWER, this approach shows a possible computational improvement for increasing number of nodes, but it presents a stronger influence on the network topology.

The impact of graph computing techniques will be even more relevant for modelling and analysing multi-energy networks. Graph computing techniques could indeed provide an approximate flow estimation without the need of solving complex algorithms combining flows on different energy carriers.