Prospective Students







2010-11 Seminars

Wind Power Forecasting

Bri-Mathias Hodge Post-Doctoral Researcher National Renewable Energy Laboratory

Wind power forecasting is an important consideration in integrating large amounts of wind power into the electricity grid. The wind power forecast error distribution assumed can have a large impact on the confidence intervals produced in wind power forecasting. We have examined the shape of the error distributions for different short-term forecasting techniques (persistence, ARMA, ARIMA) for wind plants in the ERCOT system over multiple timescales. Comparisons are made between the experimental distribution shapes and that of some commonly assumed distributions, including the normal, beta and Weibull distributions. The shape of the distribution is found to change significantly with the length of the forecasting timescale and the forecasting technique used. The Cauchy distribution is proposed as a model distribution for the forecast errors associated with the persistence model and model parameters are fitted. Non-parametric kernel density estimation techniques are applied to model the errors produced by the statistical techniques. The differences in confidence intervals obtained using the newly characterized distributions and the commonly assumed normal distribution are compared. The practical implications of these different confidence intervals are then examined through their use in a stochastic unit commitment and economic dispatch model.