Time Series modelling for solar irradiance estimation in northeast Brazil

Abstract

In this work we used time series modelling for generating synthetic sets of hourly solar irradiance for the city of Petrolina located in the northeastern region of Brazil. The models were obtained for each month and were based on 20 years of satellite data. For each month, four time series structures were investigated: auto regressive, auto regressive integrated, auto regressive moving average and auto regressive integrated moving average. We compare the 48 obtained models by comparing the the mean of the synthetic series with the data set means. Another comparison made to validate the time series model was the square error between the data histogram and the synthetic series histogram. Results show that the different model structures generate the best fitting synthetic data for the studied city. This work describes the process of pre-filtering of the data for finally obtaining the monthly models. It also presents the generated synthetic series for hourly solar irradiation. The process described in this work might be used in the planning phase of a solar farm by generating stochastic data for solar irradiance estimation.