The global need for an energetic matrix alternative to fossil fuel is urgent, therefore we need to rely on alternative energy sources, such as photovoltaic solar energy, that has been gaining broad relevance and currently stands as one of the industries that grows at highest pace in the world. Facing the difficulties regarding energy generation through renewable sources, the modelling and prediction of time series appear as a fundamental tool to reduce uncertainty. According to this context, this article presents a detailed descriptive analysis for an hourly solar power time series of a 78MWp photovoltaic power plant located in the Northeast Region of Brazil, bringing up its main features, where it is possible to highlight the double seasonality. Subsequently, based on these properties, the predictive capacity of three different statistical methods for forecasting time series are compared, in which one of them is an exponential smoothing method for the seasonal adjusted series, the other one uses a combination of Fourier terms with an exponential smoothing method and, the last one, a dynamic harmonic regression method. The MASE measure of forecast accuracy is used to evaluate the capacity of the methods and the exponential smoothing method for the seasonal adjusted series was the one that obtained better overall performance.