Time Series Forecasting for Solar Power Generation using Support Vector Regression with Extreme Gradient Boosting

In modern era, time series forecasting plays a crucial role in Solar Power Generation (SPG), by allowing the prediction of future energy output using past power and weather data for grid stability and efficient resource planning. Even though, previous researchers have developed various advanced model for time series prediction in SGP, still there are challenges such as, ineffective with non-linear data patterns and generalization. Therefore, to overcome these issues, this research proposes a hybrid model namely Support Vector Regression with Extreme Gradient Boosting (SVR-XGBoost). Initially, the input data is collected from Tubas Electricity Company and online records of NASA. After that, preprocessing is performed by using min max normalization. Furthermore, Pearson Correlation Coefficient (PCC) is performed to identify and select the most relevant features to the output variable. Lastly, a hybrid Machine Learning (ML) model namely SVR-XGBoost is employed for forecasting of time series data in SPG. The proposed hybrid model, SVR-XGBoost attained better results in the context of Mean Absolute Error (MAE) and Root Mean Square Error (RMSE) with 0.2010 and 0.2550 respectively when compared to existing Natural Gradient Boosting (NGBoost) model.