Monthly Clean Electricity Generation Forecasting in Thailand via STL, Holt, and SVR-GA

Due to the demand for electricity to drive Thailand’s economy and the awareness of environmental impacts, Thailand mainly uses natural gas and renewable energy sources for electricity generation. However, the amount of production from clean energy sources at each period is uncertain, which affects electricity production planning and policies. This research presents a hybrid model of STL decomposition, Holt, support vector regression, and genetic algorithm that emphasizes improving the quality of time series data before building a forecasting model to accurately predict electricity generation from clean energy sources. In addition, the proposed model is compared with conventional statistical forecasting models and other hybrid models to evaluate the forecasting performance using three performance criteria. The results indicate that the proposed hybrid model performs superior to other forecasting models. Furthermore, the proposed hybrid model can provide reliable and accurate forecasting results even when applied to new additional data. Therefore, the proposed hybrid model can be applied as a reliable tool for supporting strategic electricity planning and policy formulation in Thailand.