Power curve characterization I: Improving the bin method

Abstract

Neural network approaches have been used to try and characterise the power performance of wind turbines in a wind farm. Stochastic techniques are often dismissed as being inferior to those which use neural networks but those stochastic techniques proposed are often overly simplistic. We propose modifications to the IEC 61400-12 bin method in order to provide a more accurate characterisation of wind turbine performance in complex terrain taking account of direction and using a varying power law dependence of power output on wind speed. The method is validated using data from a real wind farm.