Evaluation of two simple wind power forecasting models

This paper investigates the performance of two simple wind power prediction models, an autoregressive one with exogenous input (ARX-model) and a neural network based one, none of which employs weather prediction data. The models are applied for predicting wind power production in three different wind parks, for which data are available. The error of the models is investigated for various forecasting horizons and statistical distributions of the error are calculated and presented. The performance of the two models is evaluated in comparison to the standard Persistence method. The models evaluated present a limited accuracy and their application might be possible for short forecasting horizons.