Modelo generalizado de valores extremos aplicado para irradiação solar e análise de confiabilidade de sistemas fotovoltaicos residenciais.

Abstract

Stochastic prediction procedures have been used for the past several decades to model solar irradiance. The most widely used procedures include integrated moving averages, autoregressive models, Beta and Weibull distributions to model hourly radiation. This thesis proposes a new methodology for probabilistic assessment of Photovoltaic power system production based on the Generalized Extreme Value distribution which is quite flexible and was able to model every hour in every season in Campina Grande. Furthermore, a reliability assessment utilizing Monte Carlo Methods was conducted to evaluate the performance of the hardware composing a photovoltaic System, a comparative study was conducted between a vulnerable system that can fail and another system that is always available, the difference between the failure probabilities is greater than 22%, which demonstrates the importance in considering the system availability in order to gather accurate results. Finally, the solar irradiance model and the system availability model were then combined with the IEEEReliability Test System load model to finally measure the photovoltaic system reliability when installed in Campina Grande. When comparing the capacity of power generation of the panels and the demand for electricity of a residence, an eight panels photovoltaic system is able to maintain a less than 10% failure probability between 9 a.m. and 1p.m., a fifteen panel system, on the other hand, is able to maintain a less than 10% failure probability between 9 a.m. and 2 p.m..