Análise de radiação solar usando modelos Sarima e suavização exponencial de Holt Winters

Abstract

Energy use from solar radiation can be a sustainable alternative to fossil fuel use, especially in regions where geographical aspects make energy supply difficult, such as the state of Amazonas. Many mathematical models are currently being made to make predictions of historical solar radiation series. The objective of this work is to analyze the solar radiation time series of the municipality of Humaitá-AM. The methodology used for the analysis of the solar radiation time series from 2009 to 2018 was the SARIMA and HOLT WINTERS Exponential Smoothing models, for the parameters modeling and simulation. The analysis of the series and choice of the best model was based on the error parameters: Mean Percentage Error (MAPE), Mean Absolute Error (MAE) and Mean Square Error (REMQ) in addition to the correlograms. The SARIMA model that best fitted the real series was the model (4,1,2) (2,1,2) 12. Comparing the two models through graphical analysis, it was possible to verify that the HOLT-WINTERS model perfectly analyzed the solar radiation series data, whereas the SARIMA model could not follow the beginning of the series. Comparing the error parameters between the two models, the SARIMA model presented the highest MAPE (9,740), MAE (0,365) and REMQ (0,5169) values, while the HOLT-WINTERS model presented lower values in MAPE (7,872), MAE (0.304) AND REMQ (0.456), thus showing to be the best model. It was concluded that both models presented satisfactory results for the solar radiation analysis of the municipality of Humaitá-AM. The model that best represented the series was the HOLT-WINTERS model that fit the complete series.