A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Artificial neural network based prediction model of daily global solar radiation in Morocco
In this study, an artificial neural network (ANN) daily global solar radiation (DGSR) predicting model was developed. In addition to geographical coordinates and the Julian day number, the inputs include six atmosphere physical parameters: maximal and minimal daily temperature, maximal and minimal daily humidity, daily total column of evaporating water and sunshine duration. The only output of the model is the DGSR. The used data is from eight weather stations spread out on different climate Moroccan regions, 80% of this data is chosen randomly to train the ANN while the remaining was used for the testing and validating phases. Several training algorithms are tested to minimize the sum of square error between the target and output and many combinations of the input parameters are tested to reduce their number. It is found that the ANN trained by The Broyden-Fletcher-Goldfarb-Shanno algorithm gives 0.98 as correlation, 1% as mean absolute percentage error and 1.2 MJ/m2/day as root mean square error both in training and validating phases. The trained and validated results of this model show a great accuracy either for estimating the DGSR in Moroccan regions where this parameter lack or for forecasting it in modeling process.
Artificial neural network based prediction model of daily global solar radiation in Morocco
In this study, an artificial neural network (ANN) daily global solar radiation (DGSR) predicting model was developed. In addition to geographical coordinates and the Julian day number, the inputs include six atmosphere physical parameters: maximal and minimal daily temperature, maximal and minimal daily humidity, daily total column of evaporating water and sunshine duration. The only output of the model is the DGSR. The used data is from eight weather stations spread out on different climate Moroccan regions, 80% of this data is chosen randomly to train the ANN while the remaining was used for the testing and validating phases. Several training algorithms are tested to minimize the sum of square error between the target and output and many combinations of the input parameters are tested to reduce their number. It is found that the ANN trained by The Broyden-Fletcher-Goldfarb-Shanno algorithm gives 0.98 as correlation, 1% as mean absolute percentage error and 1.2 MJ/m2/day as root mean square error both in training and validating phases. The trained and validated results of this model show a great accuracy either for estimating the DGSR in Moroccan regions where this parameter lack or for forecasting it in modeling process.
Artificial neural network based prediction model of daily global solar radiation in Morocco
Mejdoul, R. (author) / Taqi, M. (author) / Belouaggadia, N. (author)
2013-11-01
9 pages
Article (Journal)
Electronic Resource
English
| American Institute of Physics | 2012
| British Library Online Contents | 2019
New temperature-based models for reliable prediction of monthly mean daily global solar radiation
| American Institute of Physics | 2018
Prediction of daily failure rate using the serial triple diagram model and artificial neural network
| DOAJ | 2022
Artificial Neural Network Models of Daily Pan Evaporation
| British Library Online Contents | 2006