Implications of Sub-Hourly Solar Radiation Variability on Decentralized Energy Systems - Generation of Synthetic Time Series and Model-Based Assessment

Abstract

Due to the recent decrease in costs, solar based renewable energy technologies like photovoltaics (PV) are taking over the global lead in terms of annually installed capacities. With an increasing share of solar energy, the volatility of solar irradiance puts additional stress and balancing requirements on the energy system. Yearly and daily solar variability are well understood and assessed in energy system models using satellite derived time series on a temporal resolution of up to 15-min. However, short-term variability of solar irradiance induced by cloud movements can only be captured and analyzed using data of ground-based measurements on a temporal resolution of minutes. As ground-based measurements are costly and only sparsely available, a novel method to generate synthetic time series of Global Horizontal Irradiance (GHI) and Direct Normal Irradiance (DNI) is proposed in this thesis. The method uses a neural network based classification of variability classes from satellite derived cloud parameters to assemble a 1-min time series for both GHI and DNI for arbitrary locations. The model is able to accurately reproduce the typical variability characteristics (e.g distribution of ramps, KSI-test) at three Baseline Surface Radiation Network (BSRN) test sites and one validation test site. Subsequently, a PV-battery model of a simple household and a low-voltage grid model are used to demonstrate the impact of using 1-min time series data over lower resolutions on the output of decentralized energy system models. The findings are interpreted in the context of decentralized energy system modeling and general recommendations for the temporal resolution of solar irradiance input data are derived.