Wenyang Deng, Dongliang Xiao, Loi Lei Lai, Fajril Mardiansah, Haiqing Cai
Reliable dispatch in renewable-rich power systems increasingly depends on how accurately net-load uncertainty is quantified and how effectively flexible resources are deployed against tail-risk events. However, widely used mixture-based uncertainty models often suffer from artificial peak formation and poor smoothness under low-component settings, which can distort multimodal net-load distributions and propagate errors into downstream risk-aware operational decisions. To address this issue, this paper proposes an aggregated system-level risk-aware dispatch framework that couples peak-suppressed net-load uncertainty modeling with the coordinated flexibility of electric vehicles (EVs) and stationary energy storage systems (ESSs). First, a covariance-corrected, density-preserving reconstruction method is developed to suppress artificial peaks and improve the probabilistic representation of multimodal net-load behavior using substantially fewer mixture components. Then, the corrected uncertainty model is used as the scenario-generation basis for a two-stage dispatch formulation under the worst-case conditional value-at-risk (WCVaR) criterion, where EV-ESS ramping flexibility is explicitly modeled to hedge against extreme forecast deviations and support real-time corrective actions. Case studies based on real-world data show that the proposed method effectively eliminates artificial peaks, improves uncertainty modeling accuracy with an 18.7% increase in log-likelihood and a 30.9% reduction in chi-square error, and improves the trade-off between reported operating cost and tail-risk protection under the adopted system-level scheduling framework. These results suggest that, under the adopted aggregated system-level formulation and tested case setting, the proposed framework can improve risk quantification consistency and support more robust risk-aware scheduling for renewable-rich power systems.