EWMA control charts have been used by several authors to monitor residuals of power curves for fault detection in wind turbines. In this study, we analyze the effect of non-constant variance in the residuals of a power curve on the performance of an EWMA control chart. We show that the variance can be modelled as a function of wind speed and use this insight to derive wind speed-adjusted control limits. A subsequent application of this procedure to a real data set spanning 2.5 years and on simulated data shows that it is able to slightly reduce the false alarm rate of the EWMA control chart.