Dynamic feasible region aggregation for enhancing operational resilience of flexible resources

Virtual power plant dispatch is challenged by ambiguous feasible region boundaries, inadequate quantification of renewable power uncertainty, inefficient coordination of heterogeneous resources, and limited resilience under disturbances such as wind turbine tripping. To address these issues, this paper proposes an integrated dispatch framework centered on dynamic feasible region construction. First, a renewable power forecasting framework combining variable screening, complete ensemble empirical mode decomposition with adaptive noise, a physics-informed neural network, and quantile regression is developed to produce 90% prediction intervals for uncertainty characterization. Second, a multi-constraint linear programming model is established, and the time-coupled active-reactive feasible region is constructed using a bidirectional squeezing procedure together with sampling points to support feasible and efficient dispatch. Third, a hierarchical curtailment-minimizing dispatch strategy is designed to coordinate gas turbines, energy storage, and controllable loads, while a hierarchical post-fault response strategy gives priority to energy storage and controllable load regulation, with gas turbines serving as backup. Comparative results show that the proposed framework achieves high renewable-energy accommodation in the studied medium- and low-penetration scenarios, reduces total operating cost by 12.3%–15.7% compared with conventional strategies, and decreases post-fault operating losses by 28.4% after wind turbine tripping. These results indicate that the proposed framework can improve both economic performance and operational resilience for virtual power plants under uncertainty and fault conditions.