Caracterización, modelización y predicción a corto plazo de la producción de la potencia de una planta fotovoltaica, utilizando cámara de cielo y Técnica de Inteligencia Artificial

Abstract

In recent times, photovoltaic production has increased exponentially in the energy industry as an alternative source in the fight against climate change. The variability of the solar resource affects these technologies, causing uncertainty in the stage of power generation to the grid. Therefore, PV production forecasting models are a useful tool for a grid operator, which aims to maintain the reliability and quality of the power grid. The most referenced forecasting methods in the literature are machine learning techniques. These techniques derive from a sub-area of artificial intelligence and their main objective is to develop models from historical data that can solve practical problems. In general, the main meteorological variables that influence the solar resource are cloud cover and fouling of PV modules. Therefore, the objective of this doctoral thesis is to quantify and measure these phenomena to develop a methodology based on sky camera images and artificial intelligence techniques for short-term photovoltaic production forecasting.