Non-Gaussian dynamical modeling of wind power generation

Abstract

The relationship of dependence between wind speed and wind power variables shows a degree of complexity that has motivated several scientific researches over the years. Much of this research seeks to understand the stochastic nature of both phenomena. It is possible to say that the Weibull distribution is well consolidated in the literature for modeling wind speed. Thus, several studies have been proposed involving this distribution, aggregating it into different modeling strategies, either for the purpose of its marginal understanding or for the purpose of joint modeling to forecast of generation wind energy. This thesis will present the reader with two essays of analysis. The first one will be focused on marginal wind speed modeling, encompassing the simulation of scenario generation combined with generalized dynamic models through two probabilistic distributions: the Weibull distribution and the Gamma distribution. The results led to important interpretations for understanding the phenomenon. The second essay will focus on joint modeling to forecast wind energy generation based on information obtained marginally from the wind speed variable. In this stage of analysis the reader will be introduced to two approaches that incorporate the generation of scenarios in dynamic structures: the dynamic linear model and the latent Gaussian process. The results indicated promising results for the Gamma distribution and show that both approaches deserve special attention in the face of such a phenomenon.