Linking the grid: A knowledge graph approach to France’s electricity consumption

Understanding and monitoring electricity consumption is critical for ensuring the stability, efficiency, and sustainability of energy systems, particularly in the context of national energy transitions. In France, where electricity production is increasingly shaped by variable renewable sources and evolving consumption patterns, accurate and timely insights into demand dynamics are essential for both grid management and

policy-making. However, electricity usage data is inherently heterogeneous, spanning diverse spatial scales and temporal resolutions. It originates from various sources and can be combined with other relevant data such as weather data and socio-economic indicators. In this work, we present our methodology for building a very large knowledge graph reflecting the electricity consumption in France enriched with numerous other data sources. We describe the ontologies we developed and re-used, the ETL conversion process and how we align the data on a fine-grained spatial and temporal scale. Next, we experiment with various graph embeddings techniques in order to compute the similarity between small geographical areas according to their electricity consumption pattern. We demonstrate that the symbolic knowledge injected in the knowledge graph enables

us to obtain different similarities depending on macro criteria. Finally, we present an end-user web application that visualizes this very large knowledge graph while being very reactive thanks to various optimizations.