Related Experiment Video
Updated: Jul 9, 2025
10:46
A Method of Trigonometric Modelling of Seasonal Variation Demonstrated with Multiple Sclerosis Relapse Data
Published on: December 9, 2015
10.7K
Short-term solar radiation forecasting using machine learning models under different sky conditions: evaluations and
Brahim Belmahdi1, Abdelmajid El Bouardi2
1Energetics Laboratory, ETEE, Faculty of Sciences, Abdelmalek Essaadi University, Tetuan, Morocco. belmahdi.brahim@gmail.com.
Environmental Science and Pollution Research International
|November 29, 2023
Summary
The random forest (RF) model accurately forecasts short-term global solar radiation (GSR). This machine learning approach offers a reliable alternative for solar energy applications, outperforming other tested models.
Area of Science:
- Renewable Energy
- Machine Learning
- Atmospheric Science
Background:
- Accurate short-term solar radiation forecasting is vital for solar energy applications and investments.
- Solar energy offers environmental benefits over fossil fuels.
Purpose of the Study:
- To evaluate and compare seven machine learning models for forecasting hourly global solar radiation (GSR).
- To assess model performance across different times, seasons, and sky conditions.
Main Methods:
- Collected meteorological, astronomical, computational, and geographical data from 2012-2015 at two locations.
- Employed and compared seven machine learning models: MLP, FFBP, ARIMA, LR, RBFNN, RF, and GPR.
- Forecasted hourly GSR using collected data as input.
Main Results:
- The random forest (RF) model demonstrated satisfactory accuracy in forecasting GSR.
- MLP and GPR models generally outperformed LR, FFBP, RBF, and ARIMA.
- RF achieved R² values of 0.9621 (Tetuan) and 0.9534 (Tangier).
- RF, MLP, and GPR models showed a tendency to under-forecast high radiation on clear days.
Conclusions:
- The proposed RF model is a reliable and accurate alternative for short-term global solar radiation forecasting.
- Further investigation into data distribution may improve forecasting for clear sky conditions.
