Titre :	Study and forecast of solar radiation using hybrid models
Type de document :	texte imprimé
Auteurs :	DJELDJLI, Halima, Auteur
Editeur :	جامعة احمد درايعية ادرار
Année de publication :	2024
Langues :	Anglais (eng)
Résumé :	Solar irradiation prediction is essential to the planning and development of solar energy systems. especially in regions characterized by a dry climate such as southern Algeria. This thesis investigates the performance of different forecast models in predicting daily global solar radiation (GSR) across six cities in the region: Adrar, Ouargla, Béchar, El Oued, Timimoun, and Tamanrasset. Artificial neural networks (ANN), GA-ANN, and FFA-ANN—are evaluated for their accuracy and reliability. The study reveals that the hybrid prediction models we developed, FFA-ANN and GA-ANN, outperform standalone ANN models in terms of prediction accuracy across all sites. While the GA-ANN model optimizes the network weights when predicting GSR, the FFA-ANN model chooses the best combination set of inputs for each site to ensure the best performance of the model. The two models achieved very good statistical indicators results , the coefficient of determination (R=0.9076 for GA-ANN, and R=0.9321 for FFA-ANN), and the average absolute percentage error (MAPE), and relative root mean square error (rRMSE). Furthermore, the research highlights the significance of input parameters such as extraterrestrial solar irradiation (H0), declination, average temperature (Tavg), and relative humidity (RH) in enhancing the predictive capabilities of the ANN model. The findings underscore the potential of hybrid forecasting models in optimizing solar radiation forecasting and facilitating informed decision-making in solar energy system installations and sizing. Overall, this thesis contributes valuable insights into the development and application of hybrid forecasting models for solar radiation forecasting in arid climates. By leveraging advanced optimization techniques and comprehensive data analysis, the study lays the foundation for enhancing the efficiency and effectiveness of solar energy systems in southern Algeria and similar regions worldwide. solar irradiation prediction is essential to the planning and development of solar energy systems. especially in regions characterized by a dry climate such as southern Algeria. This thesis investigates the performance of different forecast models in predicting daily global solar radiation (GSR) across six cities in the region: Adrar, Ouargla, Béchar, El Oued, Timimoun, and Tamanrasset. Artificial neural networks (ANN), GA-ANN, and FFA-ANN—are evaluated for their accuracy and reliability. The study reveals that the hybrid prediction models we developed, FFA-ANN and GA-ANN, outperform standalone ANN models in terms of prediction accuracy across all sites. While the GA-ANN model optimizes the network weights when predicting GSR, the FFA-ANN model chooses the best combination set of inputs for each site to ensure the best performance of the model. The two models achieved very good statistical indicators results , the coefficient of determination (R=0.9076 for GA-ANN, and R=0.9321 for FFA-ANN), and the average absolute percentage error (MAPE), and relative root mean square error (rRMSE). Furthermore, the research highlights the significance of input parameters such as extraterrestrial solar irradiation (H0), declination, average temperature (Tavg), and relative humidity (RH) in enhancing the predictive capabilities of the ANN model. The findings underscore the potential of hybrid forecasting models in optimizing solar radiation forecasting and facilitating informed decision-making in solar energy system installations and sizing. Overall, this thesis contributes valuable insights into the development and application of hybrid forecasting models for solar radiation forecasting in arid climates. By leveraging advanced optimization techniques and comprehensive data analysis, the study lays the foundation for enhancing the efficiency and effectiveness of solar energy systems in southern Algeria and similar regions worldwide.

Study and forecast of solar radiation using hybrid models [texte imprimé] / DJELDJLI, Halima, Auteur . - جامعة احمد درايعية ادرار, 2024.
Langues : Anglais (eng)

Résumé :

Solar irradiation prediction is essential to the planning and development of solar energy systems. especially in regions characterized by a dry climate such as southern Algeria. This thesis investigates the performance of different forecast models in predicting daily global solar radiation (GSR) across six cities in the region: Adrar, Ouargla, Béchar, El Oued, Timimoun, and Tamanrasset. Artificial neural networks (ANN), GA-ANN, and FFA-ANN—are evaluated for their accuracy and reliability. The study reveals that the hybrid prediction models we developed, FFA-ANN and GA-ANN, outperform standalone ANN models in terms of prediction accuracy across all sites. While the GA-ANN model optimizes the network weights when predicting GSR, the FFA-ANN model chooses the best combination set of inputs for each site to ensure the best performance of the model. The two models achieved very good statistical indicators results , the coefficient of determination (R=0.9076 for GA-ANN, and R=0.9321 for FFA-ANN), and the average absolute percentage error (MAPE), and relative root mean square error (rRMSE). Furthermore, the research highlights the significance of input parameters such as extraterrestrial solar irradiation (H0), declination, average temperature (Tavg), and relative humidity (RH) in enhancing the predictive capabilities of the ANN model. The findings underscore the potential of hybrid forecasting models in optimizing solar radiation forecasting and facilitating informed decision-making in solar energy system installations and sizing. Overall, this thesis contributes valuable insights into the development and application of hybrid forecasting models for solar radiation forecasting in arid climates. By leveraging advanced optimization techniques and comprehensive data analysis, the study lays the foundation for enhancing the efficiency and effectiveness of solar energy systems in southern Algeria and similar regions worldwide. solar irradiation prediction is essential to the planning and development of solar energy systems. especially in regions characterized by a dry climate such as southern Algeria. This thesis investigates the performance of different forecast models in predicting daily global solar radiation (GSR) across six cities in the region: Adrar, Ouargla, Béchar, El Oued, Timimoun, and Tamanrasset. Artificial neural networks (ANN), GA-ANN, and FFA-ANN—are evaluated for their accuracy and reliability. The study reveals that the hybrid prediction models we developed, FFA-ANN and GA-ANN, outperform standalone ANN models in terms of prediction accuracy across all sites. While the GA-ANN model optimizes the network weights when predicting GSR, the FFA-ANN model chooses the best combination set of inputs for each site to ensure the best performance of the model. The two models achieved very good statistical indicators results , the coefficient of determination (R=0.9076 for GA-ANN, and R=0.9321 for FFA-ANN), and the average absolute percentage error (MAPE), and relative root mean square error (rRMSE). Furthermore, the research highlights the significance of input parameters such as extraterrestrial solar irradiation (H0), declination, average temperature (Tavg), and relative humidity (RH) in enhancing the predictive capabilities of the ANN model. The findings underscore the potential of hybrid forecasting models in optimizing solar radiation forecasting and facilitating informed decision-making in solar energy system installations and sizing. Overall, this thesis contributes valuable insights into the development and application of hybrid forecasting models for solar radiation forecasting in arid climates. By leveraging advanced optimization techniques and comprehensive data analysis, the study lays the foundation for enhancing the efficiency and effectiveness of solar energy systems in southern Algeria and similar regions worldwide.