Two-Level Ensemble Methods for Efficient Assessment and Region Visualization of Maximal Frequency Deviation Risk

This paper proposes two-level ensemble methods for an efficient risk assessment of the maximal frequency deviation (MFD) caused by stochastic outputs of the multiple renewable energy sources (RESs) in the operational planning. At the foundation level, an ensemble method (PAM-CH) is proposed to enable a fast evaluation and region visualization of MFD risk by iteratively applying partitioning around medoids (PAM) and convex hull (CH). Only the representative samples (RSs) of RES outputs selected by PAM are used to conduct a limited number of time-domain simulations (TDSs), and construct and update the convex hull-based MFD risk region (CHRR) according to the preset risk assessment matrix (RAM). Then relying on the established CHRR, an advanced-level ensemble method (DBSCAN-PAM) is further proposed to conduct an even faster MFD risk assessment to handle varying distributions and correlations of RES outputs depending on different risk assessment tasks (e.g., different planning timescales and RES types). A very limited number of outliers and RSs determined by the density-based spatial clustering of application with noises (DBSCAN) and PAM need to be assessed in this level. The accuracy and efficiency of the proposed methods are validated in a modified IEEE 68-bus benchmark system and a real provincial power grid in East China.