Predicting Solar Radiation in Iraq using principal component analysis and multiple regression model

Abstract

The goal of study to develop an effective predictive model for daily Total Global Solar Radiation (TSLR) levels at five meteorological Stations (Abu Ghraib, Al Qurnah, Balad, Tikrit, and Khanaqin) using combines of principal component analysis (PCA) and Multi linear regression (MLR) based on a fourteen factors namely RHmax, RHmin, RHavg, ET, ATmin, ATmax, ATavg, WSavg, WSmin, 2-meter dewpoint temperature, SP, CS, UV downward radiation, and CC. Five prediction models have been developed, present an intense relationship with TSLR over period 2014-2023, as evidenced by a high R (0.872 - 0.893) and an adjusted R² value of (0.761- 0.798 ) displaying a strong fit model, and MAE (2.461-3.01), MSE (4.89- 9.13), RMSE (3.02-3.53), and MAPE (16.46% - 21.23%). The validations and comparisons conducted for last year data (2024), yielded a strong R (0.871-0.889). In addition, the asset values of R2 were (0.711-0.780) for considered station respectively indicating strong fit models. The models achieve a relatively MAE ranged (1.75-2.459), MSE (3.92-7.18), RMSE (2.24-2.95), and MAPE (13.55% - 18.47%). These results show a highly accurate prediction models from closely approximate between the predicted and observed values. The validation and comparison supported research results, as they successfully revealed high accuracy of the regression models.