Hourly Irradiance Forecasting in Malaysia Using Support Vector Machine

This paper investigates the use of support vector machine (SVM) to forecast hourly solar irradiance for a tropical country. The hourly irradiance data was obtained from Sepang Malaysia, recorded throughout 2011. The data is converted into corresponding clearness index values to facilitate model convergence. The forecast is tested against the standard multilayer perceptron (MLP) technique and persistence forecast. The evaluation metrics used to validate each model's performance are mean bias error, root mean square error, mean absolute error/average, and Kolmogorov-Smirnov integral test. Results show that the SVM performs significantly better than the conventional MLP technique.