GARCH in mean type models for wind power forecasting

Abstract

Wind power penetration to power system is increasing with a very fast speed. Study on wind power forecasting is beneficial to the stable operation and economic dispatch of power systems. To improve the forecasting performance of wind power, it is necessary to investigate on the intrinsic characteristics of wind power. As the most recognized characteristics of wind power, volatility and intermittency are widely concerned. In this paper, GARCH in mean (Generalized Autoregressive Conditional Heteroskedasticity in mean) type models are presented for wind power forecasting, and the impacts of volatility and intermittency to wind power time series is modeled in the mean equation of the forecasting model. With the help of GARCH in mean effect curve, the negative impact of volatility to wind power is highlighted. By means of the Conditional Maximum Likelihood Estimation (CMLE) method, the parameters are estimated for all the proposed models. In case study, wind power forecasting based on the two types of proposed models are carried out using the historical coastal wind power data of East China. Compared with the time persistence model, Auto-regressive Moving Average (ARMA) model and GARCH model, the proposed GARCH in mean type models are validated to be effective and outperform the classical wind power forecasting models.