Statistical Analysis of Estimation of Global Solar Radiation with two Models at Uyo

Abstract

Atmospheric parameters of maximum temperature, relative humidity, sunshine hours and cloudiness for a period of seventeen years (1991-2007) were used to forecast the global solar radiation at Uyo, Nigeria with artificial neural network (ANN) and Angstrom-Prescott models. The error values Angstrom-Prescott model are MBE = (2.0725, 0.0978, 0.0685, -0.0018), RMSE = (2.1955, 0.6356, 0.7501, 0.6463), MPE = (-12.5813, -1.0589, 3.2699 3.8226), while ANN model are MBE = (0.0387, 0.0796, -0.0343, 0.0002), RMSE = (0.5341, 0.2635, 0.0468, 0.0536) and MPE = (-0.5057, 0.2901, -0.1715, 0.0518) for sunshine hours, maximum temperature, relative humidity and cloudiness respectively. The validation results of the two models showed that they have predicting capacity, but ANN has better predicting capacity. It, therefore, becomes clear that ANN has better agreement with measured global solar radiation, hence should be used for forecasting of global solar radiation at Uyo and other locations with similar climatic condition.

Keywords: Angstrom-Prescott, Artificial neural network, MBE, RMSE, MPE and correlation