Wind Turbine Flicker Calculation Using Neural Networks

Abstract

The connection of wind turbines to the distribution networks may affect the voltage quality offered to the consumers. One of the factors contributing to this effect are the rapid variations of the wind turbine output power, which cause respective fluctuations in the supply voltage, referred to as flicker. This paper presents a neural network based model for wind turbine flicker emission calculations. Neural network training patterns are developed using a simulation model of a typical 500 kW stall-controlled wind turbine, by varying all wind and network parameters that might affect the expected flicker levels. The proposed neural network model predicts flicker emissions with sufficient accuracy under any normal operating conditions (wind speed mean value and turbulence intensity) and network characteristics (short circuit capacity, angle of Thevenin impedance and local load). The paper also includes an extensive discussion on the dependence of the flicker severity on the wind and network parameters considered.