Fast Estimation of Power Supply Assurance Capacity Under Extreme Weather Uncertainty: A Transformer and Monte Carlo Hybrid Approach

In the context of uncertain renewable energy generation and power loads vulnerable to extreme weather disturbances (e.g., heatwaves, cold waves, rainstorms), traditional power supply guarantee prediction methods face challenges in handling complex variability. This paper presents a refined strategy for predicting and evaluating power supply guarantee demands by integrating an improved Transformer model with Monte Carlo simulation: a multi-quantile time series prediction model, optimized via the particle swarm optimization (PSO) algorithm to enhance generalization and accuracy, is first developed to forecast photovoltaic (PV), wind turbine (WT), and power demand (PD) outputs, with experimental results showing that the improved Transformer model outperforms conventional LSTM and CNN-LSTM models, achieving an average mean squared error (MSE) of 0.0071 and a mean absolute error (MAE) of 0.069; additionally, to address the unpredictability of renewable energy under extreme conditions, a dynamic adjustment mechanism embedded with Monte Carlo simulation generates diverse scenarios to simulate output and load fluctuations, enabling precise evaluation of power supply guarantee boundary thresholds. The proposed framework effectively predicts power supply guarantee demands in uncertain environments, providing reliable support for power system risk assessment, optimal dispatch, and secure/reliable power supply by bridging predictive modeling with scenario-based uncertainty analysis.