Measured maximum temperature, relative humidity, cloudiness and sunshine duration measurements between 1991 and 2007 for warri, Delta state of Nigeria were used for the estimation of monthly average mean global solar radiation on horizontal surface using Artificial Neural Network and ANGSTROM-PRESCOTT model technique. This study was based on Multi Layer Perceptron (MLP) which trained and tested using past seventeen years (1991- 2007) meteorological data. The chosen weather data were divided into two randomly selected groups, the training group, corresponding to 66.7% of the patterns, and the test group, corresponding to 8.3% of patterns; so that the generalization capacity of network could be checked after training phase. Also three random months were selected as holdout data and it corresponds to 25.0%. Coefficients of determination R2 for the MLP models 0.958 indicating reasonably strong correlation between estimation and measured values. The values of RMSE for empirical model is 3.39118 indicating higher errors and low prediction and the value for MLP is 0.050106 indicating lower errors and higher prediction accuracy. Also the values of MPE for empirical model and MLP model confirms the evaluation of RMSE in the prediction of solar radiation which are -80-3875 for the former and -84.3124 for the later. The summation of MBE values were found to be -0.3675 for empirical and 0.0625 for MLP. The negative sign indicate under estimation while 0.0625 for MLP indicates low error since the actual values are positive. Figure 2 and 3 shows the graphs comparing the measured and predicted values of the two models of the monthly average mean solar radiation for warri. The result clearly show that there is a good and strong agreement between the MLP predicted and the measured values compared to empirical predicted model and the measured values. Hence comparison between the ANN model and ANGSTROM-PRESCOTT empirical models has shown the superiority of the ANN model in the prediction.
All Published work is licensed under a Creative Commons Attribution 4.0 International License
Copyright © 2019 All rights reserved. iMedPub LTD Last revised : August 21, 2019