Support Vector Regression Approach for Wind Forecasting

Abstract

The government policies to fully support the G20 sustainability emphasize increased use of renewable energy. The high penetration of wind energy into power systems poses many challenges for energy system operators, primarily due to the unpredictability and variability of wind energy production. Wind power may not be provided, but accurate forecasting of wind speed and power generation helps grid operators reduce the risk of reduced electricity reliability. Accurately predicting wind speeds over 1 to 24 hours based on these conditions is important for predicting potential energy supply. These short-term forecasts are important to support wind power planning, so the required base load supply for the grid is always guaranteed (even if the wind power output fluctuates significantly). This task demonstrates that the relative forecasting performance of a support vector regression (SVR) wind forecasting system can be improved by systematically selecting and combining related input functions that affect wind speed. Shows the results of data collected in Sidrap, Indonesia, during the six months of 2019. This paper explained key methods of wind forecasting, based on the evaluation of wind speeds and wind speed prediction methods. The RMSE from the SVR shows an 8% - 9% improvement on the RMSE of the persistence forecast every 1 hour. Wind speed estimation using a support vector regression approach has the potential for further development, one of which is determining the potential location of wind-based renewable sources and Wind Energy Conversion System (WECS) can make more efficient.