Forecasting of Solar Power Volatility using GJR-GARCH method

The rapid increase in solar power plant installation creating considerable difficulties for power system operation and control, due to the highly stochastic nature of solar energy. To ensure a smooth solar panel output, an accurate short-time forecast plays a crucial role in grid energy management. In this paper, a problem of hours ahead solar plant power prediction and forecasting is considered. This accurate prediction can significantly improve the electricity cost by planning the extra power generation for supplying loads. For a case study, a meteorological dataset for a real operated solar power plant is used to develop and verify the forecasting model. A detailed time series analysis of the solar power output (per second data) has been used for model development and validation. Additionally, after investigating different techniques, a Glosten-Jagannathan-Runkle Generalized Auto Regressive Conditional Heteroskedasticity (GJR-GARCH) based forecasting has been proposed. The developed model predicts solar power with high accuracy with a confidence interval of 99%. The results and verification with a real data set prove the capability of applying this method in industrial applications. These predicted models are utilized to generate battery control signals ensuring a smooth power flow to the grid or directly send to the grid for ensuring peak energy savings.