Impact of Averaging Period on a Medium-Scale Wind Turbine Power Curve and Prediction of Annual Energy Production

Abstract

A reliable method for the prediction of Annual Energy Production is necessary for the United States to achieve its goal of 20% wind energy by 2030. Along with the pressure to reduce error in energy yield prediction is the constant pressure for manufacturers to reduce expenses, such as the expense of a measurement campaign for a turbine. Shorter averaging periods mean shorter and less expensive campaigns. However, caution must be used to ensure that the averaging period is sufficiently long enough to account for the separation between where wind speed is measured and where the wind turbine's rotor is located. Averaging periods that are too short could mean a reduction in the correlation between wind speed and electrical power output measurements. The comparison of power curves constructed with various averaging periods indicated that shorter averaging periods have little overall impact on the general shape of the power curve. Wind turbines spend most of their time in the lower part of Region 2, where there is especially little impact on the shape of the power curve. In the higher part of Region 2 and in Region 3, there is more variation in the shape of the curve, possibly due to a decrease in the number of observations at these wind speeds. In the approximately linear range of wind speeds, correlation between wind speed and power output was significantly weakened from a 10-minute averaging period to a 1-minute averaging period. The calculation of AEP values with power curves of various averaging periods indicated that shortening to a 1-minute averaging period has much more of an impact on energy yield prediction than shortening to a 5-minute averaging period. These AEP values were calculated with a wind speed probability distribution based on a 10-minute averaging period, which is very common in meteorological data sets. If a manufacturer wants to reduce the averaging period of its wind turbine power curve to 1-minute, the impact of combining this power curve with a wind speed probability distribution of a different averaging period must be considered. This impact appears to be minor, but to reduce error in energy yield prediction, it is recommended that the averaging periods match. Caution should be advised when shortening the averaging period to a length of 1-minute. Though the standard for large turbines is a 10-minute averaging period, our results show that a 5-minute averaging period for medium-scale wind turbines doesn't have a significant impact on energy yield prediction. Switching to a 5-minute averaging period could potentially cut the length and cost of a measurement campaign in half. We hope that the methods and code presented in this paper can be easily implemented by those interested in assessing the impact of averaging period on turbine power curves and energy yield prediction.