Search

• A partial shading emulator for module-level emulations of utility-scale PV systems is proposed. • A fully customizable cloud shadow model that can reproduce real cloud coverage natures is introduced. • Two endogenous factors, namely PV array arrangement and system orientation are evaluated. • Three exogenous factors, namely shadow intensity, shadow velocity, and shadow size are explored. • Cloud shadow transitions exploited from real measurements are applied for assessing ramp-rates in reality. The short-term power output variability of solar photovoltaic (PV) systems caused by passing clouds is becoming a major concern for grid operators. As the penetration of utility-scale PV systems boosts, the rapid power fluctuations greatly challenge the grid transient stability. In this regard, the concept of power ramp-rate (RR) has been recently imposed to quantify the PV fluctuations, and a series of ramp regulations are implemented by utilities. However, due to the limitations in terms of high-resolution simulation of utility-scale PV systems and reproducing cloud shadow natures, current studies still show deficiencies in comprehensively investigating the RRs rising from both endogenous and exogenous factors during cloud shadow transitions. With the objective of producing high-resolution and high-accuracy emulations of utility-scale PV systems under passing clouds, thus providing a clearer understanding on the cloud-induced RRs, this paper first sets forth a partial shading emulator that is capable of efficiently mimicking the behaviors of an arbitrary partially shaded PV system, in accuracy of the PV module level. Then a fully customizable shadow model that can reproduce the natures of a real cloud shadow is introduced. Based on the developed emulation tools, the effects of two endogenous factors i.e., PV array arrangement and system orientation, and three exogenous factors i.e., shadow intensity, shadow velocity, and shadow size on RRs are explored for a range of utility-scale PV systems from 1 MW to 60 MW. Furthermore, in order to assess the RRs in reality, a total of 3747 cloud shadow transitions exploited from real measurements have been applied for emulations. The results reveal that the RRs caused by passing clouds are critical problems for system operations, and a larger system can suffer more from ramp violations, indicating that the advanced ramp control strategies should be essential for contemporary utility-scale PV systems. [ABSTRACT FROM AUTHOR]

• A concentric sensor network for PV power nowcasting. • A comprehensive framework for spatial and temporal predictor preselection. • Consistent PV nowcasts at a fixed prediction horizon, with cloud dynamics contained. • Strong model adaptability to various weather conditions. • A practice of integrating the nowcasts to PV grid-friendly control. The increasing penetration of photovoltaics (PV) systems introduces more uncertainties to the power system, and has drawn serious concern for maintaining the grid stability. Consequently, the PV power grid-friendly control (GFC) has been imposed by utilities to provide additional flexibilities for power system operations. Conventional GFC strategies show limitations to estimate real-time maximum available power, especially when fast moving clouds occur. In this regards, the spatio-temporal (ST) PV nowcasting using a sensor network provides a remedy to the above issue. However, current ST nowcasting methods suffer from the problems such as predictor mis-selection, inconsistent nowcasting, and poor model adaptability, which still hinder their practical use for GFC. In this paper, a novel ST PV power nowcasting method with predictor preselection is presented, which can be used for GFC. The proposed method enables a fast and precise predictor preselection in different scenarios, and provides consistent PV nowcasts with cloud information interpolated. The effectiveness of the proposed nowcasting method is evaluated in a real sensor network. The experimental results reveal that the proposed method has strong robustness in case of various weather conditions, with fewer training data used. Compared with the conventional methods, the proposed method shows an average nRMSE and nPMAE improvements over 13.5 % and 41.3 % respectively in the cloudy days. A practice of integrating the proposed nowcasting method to GFC operation is also demonstrated. The results show that the proposed method is promising to improve the performance of GFC. [ABSTRACT FROM AUTHOR]