A New Approach for Estimating Monthly Global Solar Irradiation Based on Empirical and Artificial Neural Networks Models: A Case Study of Al-Hoceima Province, in Morocco North Region

Abstract

Knowledge of the monthly global solar irradiation is crucial for the planning, design, and dimensioning of solar photovoltaic systems. However, global horizontal solar irradiation is not readily available at all sites. In this regard, the authors proposed a hybrid model that combines an empirical model with an artificial neural network technique to estimate the monthly global solar irradiation for the case study of the city of Al-Hoceima, Morocco. To this end, 17 empirical models are calibrated and evaluated in order to obtain the best empirical model, where the output of this empirical model is combined with one or more parameters such as sunshine duration (S), extraterrestrial radiation (Ho), mean, maximum and minimum temperatures (Tmean), (Tmax), (Tmin), the ratio (Tmin/Tmax), and the difference (Tmax-Tmin). These combinations of parameters are used as inputs to different MLP models that were developed and evaluated in order to find the model with the highest overall performance. The simulation results demonstrated that the proposed hybrid approach provides the best results in terms of RMSE, R2, and MAE and outperforms MLP or empirical models used individually. Also, the results indicated that the proposed model gives higher performance than five machine learning algorithms including SVR, Random Forest, Xgboost, decision tree, and K-nearest neighbors.