Modelamiento de patrones diarios de producción solar fotovoltaica orientada a la mejora de la eficiencia y precisión en las proyecciones energéticas: un enfoque estadístico

Abstract

This research develops a statistical model to characterize and predict photovoltaic (PV) solar energy production, considering its annual and daily variations based on data from three groups of solar panels over one year. To model annual variability, autoregressive moving average (ARMA) models were applied to detrended time series data, allowing the identification of temporal dependency patterns in energy generation. At the daily level, the work focused on fitting beta distributions to percentiles of photovoltaic production, and on clustering the daily profiles, which provided a refined characterization of hourly generation patterns. Furthermore, based on the obtained ARMA models, synthetic time series with statistical properties equivalent to the original series were generated, facilitating simulation and predictive analysis. The methods implemented in this study provide a robust tool to improve the accuracy of solar generation projections and optimize the management of photovoltaic systems, thereby contributing to the efficient planning and operation of renewable energy.