Statistical modeling for global solar radiation forecasting in Bogotá

Based on measurements of global solar radiation carried out in the National University of Colombia (located in Bogotá, at 74° 4´ West, 4° 35´ North, 2580 m altitude) for the period 2003-2009, a statistical modeling based on time series was made to forecast the accumulated mean daily global solar radiation in Bogotá. This analysis was done using the programs EXCEL 2007 and SPSS 16; the lost data were determined with the technique of lineal trend. Using the test of Dickey - Fuller it was established that the time series are not stationary, but seasonal. Using the methodology of Box and Jenkins to convert the original data to a stationary series, it was found that the series which better fitted to the one built with experimentally measured data, can be obtained using the ARIMA (1, 0, 0) model, which can be mathematically represented by the equation X_t = 3562.78 + 0.518 X_t-1 + ε_t, where the error ε_t is distributed like a normal (0, 1) white noise.