Enhancing Support Vector Machine Prediction Accuracy for Global Solar Radiation Modeling Using Particulate Matter

The amount of particulate matter in the surrounding can significantly affect the amount of incoming solar radiation. In this study, particulate matter [PM10 and PM2.5] is used as an additional input factor for a Support Vector Machine (SVM) model developed to calculate global solar radiation on the surface of the earth. SVM-1-x and SVM-2-x model accuracy are evaluated and compared. The comparative analysis shows that the performance of SVR-2-x models with an additional input factor, namely particulate matter, outperforms the SVR-1-x model. The percentage improvement in the SVM-2-4 model for different metrics is 15.26% for MAE, 27.87% for MSE, 14.57% RMSE and 2.63% for R2. The inclusion of PM2.5 and PM10 with other meteorological parameters such has relative humidity, minimum and maximum temperature provide more comprehensive representation of the atmospheric condition thereby enhancing the predictive capabilities of the SVR model for global solar radiation prediction.