Generation of Short-term Wind Power Scenarios from an Ensemble of Hourly Wind Speed Forecasts

Abstract

This paper proposes a methodology for generating synthetic wind speed and wind power scenarios obtained from hourly or half-hourly wind speed forecasts (e.g., made available by Brazilian Electric System Operator) and from measured or reanalysis data. It is assumed that at each hour of the study horizon the wind speed predictions are realizations of a Weibull distribution. The generation of wind speed scenarios is performed through simulations of Weibull distributions at each hour of the horizon, fitted by generalized additive models for location, scale and shape (GAMLSS), but preserving the hourly autocorrelations through the Nataf transformation. Finally, power curves transform wind speed scenarios into wind power production scenarios. The proposed approach was applied to two wind farms in the Brazilian electrical system. The results obtained showed that the proposed methodology was effective, and the results are encouraging for the application of the proposed methodology to other wind farms.