Enhancing Congestion Forecast for TSOs: a Probabilistic Approach

(PCF) approach designed to enhance grid security analyses within the studied TSO’s internal grid. It aims to improve the reliability and cost-efficiency of grid operations by accounting for key uncertainty drivers (cross-border flows, wind generation, and DSO connection points).

The main method involves transitioning from point forecasts to probabilistic forecasts using artificial intelligence (AI) techniques. The process starts with supervised learning to forecast nodal power uncertainties, followed by generating grid scenarios that consider nodal correlations using a robust statistical methodology. Congestion profiles are obtained by running an accelerated AC load-flow analysis for each scenario and contingency case. This yields a probabilistic congestion profile, offering operators insights into congestion likelihood and severity through a dynamic dashboard. A dynamic reliability margin, which accounts for the uncertainty on the predicted loading depending on exogenous factors (weather, market situations, etc.), aims at raising less false alarms when the uncertainty is low, and capturing more congestion cases when the uncertainty is large. Results show that, when models are calibrated to a fixed false-alarm rate (precision = 0.6), the probabilistic approach reduces missed congestion cases by 63%. Conversely, when calibrated to a fixed security level (10% missed congestion cases), false alarms are reduced by 47%. These findings demonstrate the value of explicitly accounting for forecasting uncertainty in congestion management. ​