Meteorological-Electrical Integrated Real-Time Resilience Assessment for Power Systems Based on Deep Learning Methods

Recent widespread power outages caused by extreme weather events have elevated expectations for power system resilience, thereby emphasizing the critical need for proactive resilience assessment. However, given the uncertain pathways and intensities of such events, along with the extensive range of potential failure scenarios, achieving real-time resilience assessment remains challenging. This paper proposes a meteorological-electrical integrated real-time resilience assessment method for power systems to efficiently quantify the resilience level and precisely identify vulnerable areas against imminent typhoon disasters. Firstly, a novel synergistic downscaling approach for typhoon nowcasting models is proposed. This method combines deep transposed convolutional neural networks and statistical downscaling approaches to provide wind speed forecasts with higher spatial resolution. Spatiotemporal mismatches between typhoon nowcasting models and power systems are then addressed through geographical alignment and multi-timescale coordination. Based on the aligned wind speeds and component fragility models, an edge-adapting, physics-informed ChebNet-based deep learning method is developed to attain efficient and accurate resilience assessment of power systems. This method calculates system resilience metrics and component vulnerability indexes throughout the event, enabling vulnerable area identification. Numerical simulations are conducted on the IEEE 39-bus system to demonstrate the effectiveness of the proposed method through comparisons.