Resumen de Synthetic generation of high-temporal resolution direct normal irradiation time series
Miguel Javier Larrañeta Gomez Caminero
In this doctoral thesis, we present two methods for the synthetic generation of high temporal resolution Direct Normal Irradiance (DNI) time series from hourly means that can be applied globally without any local adaptation.
The SA (Stochastic Adaptation) method, based on the concept proposed by Polo et al, (2011) with several improvements. It divides the irradiance into a deterministic and a stochastic component, i.e., the contribution from the hourly mean and the stochastic fluctuation which depends on the sky condition, respectively. The improvements are focused on the characterization and reproduction of the cloud transients’ effects depending on the sky condition, and the identification of periods where fluctuations of the solar radiation take place.
The ND (Non-Dimensional) method relies on the normalization of the daily profiles of an extensive database of solar radiation data by a clear-sky envelope that can be used to reproduce the high-frequency dynamics of DNI in any location. The method is based on the concept proposed by Fernández-Peruchena et al, (2015) but with several improvements, the most relevant being the identification of the most similar day to be reproduced not only in terms of energy, but also in terms of distribution and variability of the daily DNI curves.
It is worth highlighting that the methods presented in this work use measurements from one location to characterize the cloud transients and generate synthetic high resolution DNI data in any location where hourly DNI data is available, without any local adaptation.
The models have been applied in three locations with different climatic conditions. In the analysis of the performance of each model, the measured and synthetic time series have been evaluated in terms of energy, frequency distribution and autocorrelation. In terms of energy, we obtain annual differences lower than 0.2%. The similitude between measured and generated DNI distributions has been evaluated through the Kolmogorov-Smirnov test Integral (KSI), and the performance of the synthetic series when used for the evaluation of the thermal power produced by a parabolic trough (PT) plant has been assessed using the daily normalized root mean square deviations (NRMSD). The autocorrelation have been addressed by evaluating the ramp rates of both measured and synthetic data sets, The generation methods provide, for an annual 1-min synthetic data set, KSI values of ~3.3 W/m2 and ~12.9 W/m2 (depending on the generation method used), and daily NRMSD of ~0.9% and ~3.4%, respectively. For ramp rates greater than 500 W/m2۰min, the differences in the measured and synthetic sets are almost negligible. Sites selected for validating these methods are located at different climates and latitudes, suggesting their global applicability.
