Data-Driven Real-Time Congestion Forecasting and Relief With High Renewable Energy Penetration

Power grid congestion has become a recurring issue due to system uncertainties caused by the increasing integration of renewable energy sources. This study proposes a data-driven approach for intrahour congestion forecasting and proactive relief in power grids. By leveraging historical congestion data, a histogram-based gradient tree boosting (HGTB) model is trained to predict real-time congestion severity and probability using grid measurements. This enables operators to anticipate and prevent congestion events. The proposed framework integrates a proactive congestion relief strategy, incorporating a multitime-scale marking method. It facilitates coordinated collaboration among optimization control schemes and harnesses the benefits of adjustable resources with varying response characteristics. Case studies on the IEEE 118-bus power system and the China Central Region power system validate the effectiveness of the approach in accurately forecasting and mitigating congestion events. The results demonstrate precise congestion forecasts and timely preventive actions, leading to cost reduction.