Obtaining high-fidelity information on extreme photovoltaic (PV) power is critical for electric utility system planning and operations. However, a scarcity of extreme data has previously made achieving an accurate estimate of extreme PV power an intractable challenge. In response to this challenge, this paper presents Extreme PV Power Analytics (EPVA). It utilizes k-means clustering to determine which PV systems have similar behaviors in their extreme capacity factors (ECFs) in order to incorporate more extreme data in an extreme value analysis. This extreme value analysis is subsequently applied to obtain the distribution of ECFs. Zone partitioning results and ECF distribution results for The United Illuminating Company service territory are presented to validate the effectiveness and efficacy of EPVA.

Given the progressively deeper integration of distributed energy resources (DERs), evaluating the potential unintentional islanding hazards in distribution networks becomes increasingly important for distribution system planning and operations. In this paper, a rigorous theoretical analysis is used to devise a DER-driven nondetection zone (D$^{2}$NDZ) method, which is then implemented through a data-driven learning-based approach. Test results indicate that D$^{2}$NDZ can quickly and effectively estimate the nondetection zones for any given distribution feeders, while avoiding numerous and time-consuming electromagnetic transient simulations. D$^{2}$NDZ software has been deployed in Eversource Energy, a major power utility company in the northeastern U.S. In practice, D$^{2}$NDZ reduces utilities engineers’ case study time from months to just a few minutes.

This paper proposes a novel approach based on k-means clustering and extreme value theory (EVT) to spatiotemporally analyze photovoltaic (PV) extreme capacity factor (ECF). Through correlation coefficient analysis, the effects of meteorological factors on PV output are quantified into different weights. These weights are then used in a k-means clustering solver to partition the utility service territory into k geographical zones such that PV systems within each individual zone will behave similarly in terms of peak capacity factors. The processes involved are presented in great detail such that the correlation coefficients between PV output and meteorological variables are calculated; weights and normalized meteorological variables are calculated; representative PV and weather data are selected; and the value of k is determined. Extreme value theory is subsequently utilized to obtain the probabilistic distribution of the ECFs for PV systems located in a specific zone within a specific time interval. A case study based on the PV and weather data in the State of Connecticut is presented to validate the effectiveness and efficiency of the proposed approach.

Efficiently calculating non-detection zones (NDZ) becomes increasingly important when evaluating unintentional islanding risks of distribution grids that are highly integrated with distributed energy resources (DER). In this paper, a rigorous theoretical method, the DER-Driven Non-Detection Zone (D-2 NDZ), is presented to estimate NDZ for any given distribution feeders. Numerical examples indicate that by using D-2 NDZ, NDZ can be quickly and effectively obtained while avoiding numerous and time-consuming electromagnetic transient simulations. Therefore, D-2 NDZ offers utilities engineers a powerful tool to better understand and operate their systems.