Large-Scale Inference of Geo-Referenced Power Distribution Grids Using Open Data

Abstract

Power distribution grids host an increasing amount of distributed renewable generators, electric vehicles, and heat pumps worldwide. Distribution grids, however, were not designed with the goal of incorporating large shares of these technologies. These soaring challenges demand accurate and realistic grid models to assess the need for operation strategies and reinforcements that ensure reliable and economic management. Nevertheless, real models are often unavailable due to privacy and security concerns or a lack of digitized data from distribution system operators. To address this issue, we present a framework for large-scale inference of geo-referenced low- and medium-voltage grid models using publicly accessible information on power demand and transport infrastructure. First, we develop a clustering algorithm, which detects load areas served by distribution grids. Then, we obtain the graphical grid layout, i.e., a graph with the street and pathway geometries and the load point connections inside the load area. Next, we introduce a selection method for line types that assigns cost-effective conductors to grid lines while ensuring operational constraints. We demonstrate the effectiveness of our approach by inferring all the low- and medium-voltage infrastructure in Switzerland. Remarkably, the inferred grids present overall power requirements and line lengths statistically aligned with reference grids.