Ramp Rate Exaggeration during Wind Speed to Wind Power Conversion

Abstract

Wind turbine power curves are frequently used within the wind power industry and academic research to convert wind speed to wind power. In assessing power curve skill, the focus is often placed on a power curve's ability to model mean power as opposed to variability. This paper aims to assess the skill of a number of power curves - including regular curve fitting, machine learning techniques and the IEC recommended binning method - in modeling wind power variability for both a single turbine and a wind farm. It was found that all power curve methodologies tested yielded high accuracy in modeling mean wind power over long periods. However, a marked exaggeration in wind power ramps was observed in the modelled time series - typically on the order of 10-14% for a single turbine and 25-60% for a wind farm. It was found that with post-processing, such as 11 trend filtering, observed ramp exaggerations can be minimized. The findings of this study should be considered by researchers working on wind power variability assessments, forecasting and ramp predictions, so as not to overestimate wind power variability.