Time Series Sampling Rates for Wind Energy Forecasting
Idris A. Eckley1, Guy P. Nason2
1Department of Mathematics and Statistics, Lancaster University, Lancaster, United Kingdom; 2Department of Mathematics, University of Bristol, Bristol, United Kingdom

With the increased adoption of wind energy within our energy networks, there is a growing requirement to provide accurate short-term forecasts of a given wind farm's output. Due to the inherent dependence between wind power and wind speed, recent applied work has considered the development of established time series models to forecast short-term wind speeds at a site of interest. Whilst modern SCADA systems are capable of recording wind speeds on a second by second basis, typically the time series models proposed to date use either hourly or 10-minute average wind speeds. In discretising the data to this extent, it is possible that the resulting time series become aliased. In other words, the second-order structure used to forecast future wind speeds could be being misrepresented. This talk will introduce a recently proposed test of aliasing and consider its application to identify appropriate sampling rates for short-term wind speed modelling.

Keywords: Time series; Aliasing; Wavelets; Wind energy

Biography: Idris Eckley is a Lecturer in Statistics at Lancaster University. Prior to his move to Lancaster he worked for several years as a Consultant Statistician within Shell. He is the Deputy Director of the STOR-i Centre for Doctoral Training and a Visiting Fellow of the Statistics group at the University of Bristol.