Analysis of Wind Farm Output: Estimation of Volatility Using High-Frequency Data

In the financial year 2011–2012, wind farms supplied 26 % of South Australia’s electricity demand according to the Australian Energy Market Operator’s report. This contribution has risen from zero in 2003. The operation of the electricity grid depends heavily on knowledge of the variability of supply. Wind farm output displays similar conditional volatility as financial market variables. In this paper, a new method of estimating wind farm output volatility on a 5-min time scale is developed through the use of higher-frequency wind farm output data. First, an autoregressive model for the high-frequency data is developed, and it is used to derive a volatility measure for 5-min data. The results are also true in certain general situations when the high-frequency data follow an autoregressive moving average process or exhibits long memory features. The methods described here are analogous to realised volatility measures used in financial series, except that wind farm output data are measured at uniform intervals, unlike random trading times for financial transactions.