Probabilistic geo-referenced grid modeling: A Bayesian approach for integrating available system measurements

With the ongoing implementation of new climate targets, power distribution grids are increasingly integrating behind-the-meter photovoltaic (PV) systems, electric vehicle (EV) home chargers, and heat pumps (HPs). The integration of these solutions can often result in grid congestion issues, requiring appropriate mitigation measures. Designing these measures can be challenging in the absence of a digital and up-to-date model of the existing infrastructure, which is often the case at the low-voltage (LV) level. In this work, we introduce a novel two-stage Bayesian approach for establishing a probability distribution of geo-referenced power flow (PF)-ready grid models using available system measurements. We demonstrate the proposed approach in a residential region in Schutterwald, Germany. We find that integrating available system measurements can effectively enhance the quality of the distribution, yielding potential grid models that more accurately align with the existing infrastructure. Moreover, we showcase the practical utility of the proposed approach for assessing overvoltage within a specific grid segment subject to high rooftop PV adoption. While state-of-the-art baselines either fail to identify any overvoltage issues or are inconclusive, integrating available system measurements using the proposed approach offers a more reliable assessment.