Qualification of wind power forecasts

Abstract In order to manage the remarkable share of wind power as present in several utilities forecast information is needed. In the last years several respective procedures have been developed and put into operation. Up to now the outcome of the forecast tools is mainly restricted to the forecasted value itself and general information of the overall error of the procedure as e.g. the standard deviation. However, for the handling of these forecast information in the framework of e.g. power station dispatch schemes, more detailed information of the structure of the expected errors- beyond the expected standard deviation- seem to be desirable. Due to the fact that the output of wind turbine systems is limited between zero and the maximum power, the error statistics cannot follow a normal distribution. Thus, for the assessment of the probability of occurrence of a certain forecast error a model for the distribution function of the errors has to be set up. We have analysed the errors of the forecast model PREVIENTO as applied for an ensemble of installations representing the lumped power output of the turbines within a given region. Given bias free forecasts, the applicability of various models for the distribution function of the set of errors has been tested. It turned out, that the use of a beta function is justifiable for this task with respect to chi-squared tests. Together with an empirically derived parametric model for the expected standard deviation of the ensemble forecast, the knowledge of the respective distribution function allows for the assignment of risk figures to any decision taken based on the wind power forecast