Wind energy as renewable energy will be crucial in ensuring future energy supply. Foundation for agood implementation of wind power output in the electrical grid are wind forecasts. Especially windpower ramp forecasting is important for an effective generation of wind power energy by wind parks.Wind power ramps are large changes in wind power output over a relative short amount of time. Thiswork compares two statistical definitions for four Swedish wind parks using 15 years of data from theNew European Wind Atlas data. This model provides, among other variables, the wind speed and direction at for wind power relevant height with a temporal solution of 30 min. Compared to absolutedefinitions, commonly used to define wind power ramps, a statistical definition is beneficial since it considers each site’s climatology. Based on this definition, a random forest classifier identifies wind speedand direction as the most important variables when forecasting wind power ramps. When fine-tuned, therandom forest classifier could become a valuable tool for forecasting wind power ramps.