"IEEE 39, IEEE 118 and a real world provincial power system cascading failure dataset"

"An AC-based cascading failure simulation model (AC-CFM) is employed to generate simulation data, capturing the power system\u2019s dynamic behavior and fault propagation patterns under varying renewable energy penetration and load fluctuations. The simulation process starts with adjusting grid parameters (including renewable energy integration node configurations and load fluctuation levels), then enters the cascading failure processing phase: identifying all islanded subsystems, sequentially applying protection measures to each island for stability, evaluating cascading failure containment, and implementing additional protection measures if propagation continues\u2014repeating for all islands until the entire network stabilizes, finally outputting comprehensive datasets.\u200bLoad variation follows a normal distribution, with the mean set as the annual peak load (serving as the expected value) and a standard deviation of 10% of this expected value. Renewable energy penetration is configured with wind power sampled via a Weibull distribution and solar power via a Beta distribution, with an overall penetration rate of 20%. For initial failure setting: in a network with N nodes, k transmission lines (k < N) are randomly selected as initial failure lines; the scale of initial cascading failure follows a Zipf distribution (parameter p determines failure size). Under normal conditions, p is 3, while it decreases to around 2.2 during extreme weather events (leading to larger failure scales); p is fixed at 2.2 throughout simulations to ensure sufficient cascading failure data collection.\u200bThe dataset includes 10,000 total samples, split into training, validation, and test sets at a 60%:20%:20% ratio. The average number of initial contingencies ranges from 2.1 to 2.6 branch outages per scenario (consistent with realistic power system disturbance scales), and the failure distributions exhibit severe class imbalance\u2014with an average of 2.2\u20132.7 bus failures and 3.3\u20135.5 branch failures per scenario."