Progrès de la recherche en sciences appliquées Libre accès

Abstrait

Application of artificial neural network and Angstrom-Prescott models in prediction of global solar radiation of Uyo City, Nigeria with atmospheric parameters

1Ibeh G. F, Agbo G. A and Rabia, S

Prediction of monthly mean global solar radiation (GSR) based on atmospheric parameters, using Multi-layer perceptron (MLP) neural networks and an Angstrom-Prescott model has been studied. Monthly mean maximum temperature, relative humidity, sunshine hours, and cloud cover values between 1991 to 2007 for Uyo city-Nigeria (latitude 5o N, Longitude 5o E), were used in this study. The statistical analysis, Angstrom-Prescott model show that MBE=-2.05667, RMSE=4.17173 and MPE=15.4617 while that of ANN model are MBE=-0.12667, RMSE=0.26243 and MPE=0.69231. The low values of ANN model indicates reasonably strong correlation between ANN models prediction and measured values of global solar radiation for Uyo-Nigeria. The value R2 of ANN model show that 97.2% of is correct while R2 of Angstrom-Prescott model show that 74.5% is correct, therefore, ANN model is a better model for this prediction.