Akademik Veri Yönetim Sistemi
INSAC International Researches Congress on Natural and Engineering Sciences (INSAC-IRNES'23), Konya, Türkiye, 18 - 19 Mart 2023, ss.212-224
Solar radiation forecasting is of great importance in generating electricity from solar panels, growing and developing crops, establishing local weather models, and making climate forecasts. In this study, the least square-support vector machine (LSSVM) with particle swarm optimization (PSO) and variational mode decomposition (VMD) techniques was integrated for the estimation of monthly solar radiation data in Hakkâri. Historical data was used to estimate solar radiation data. Delayed solar radiation data exceeding the confidence limit of the autocorrelation function were used in the setup of the model. The data was divided into 80% training and 20% testing during the modeling phase. Model performance was evaluated using scatter diagrams and statistical criteria such as correlation and error values. As a result of the study, it has been determined that the VMD-PSO-LSSVM model is more successful than the PSO-LS-SVM model. In addition, it has been revealed that the predictive power of the PSO-LS-SVM model can be increased by using the input variables allocated to the intrinsic mode function and residuals with VMD. The study outputs can help optimize the operation of solar power plants, plan irrigation
and fertilization of farmers, improve the weather forecast, and increase the forecast performance of climate models.