Vertical wind speed extrapolation using regularized extreme learning machine

Abstract

The cost of measuring wind speed (WS) increases significantly with mast heights. Therefore, it is required to have a method to estimate WS at hub height without the need to use measuring masts. This paper examines using the Regularized Extreme Learning Machine (RELM) to extrapolate WS at higher altitudes based on measurements at lower heights. The RELM uses measured WS at heights 10-40 m to estimate WS at 50 m. The estimation results of 50 m are further used along with the measured WS at 10-40 to estimate WS at 60 m. This procedure continues until the estimation of 180 m. The RELM's performance is compared with the regression tree (RegTree) method and the standard 1/7 Power Law. The proposed algorithm provides an economical method to find wind speed at hub height and, consequently, the potential wind energy that can be generated from turbines installed at hub height based on measurements taken at much lower heights. Moreover, these methods' extrapolated values are compared with the actual measured values using the LiDAR system. The mean absolute percentage error (MAPE) between extrapolated and measured WS at the height of 180 m using measurements at the height of 10-40 m using RELM, RegTree, 1/7 Power Law, and Power Law with adaptive coefficients is 13.36%, 16.76%, 33.50%, and 15.73%, respectively.

Keywords

• wind speed
• vertical extrapolation
• regularized extreme learning machine
• regression tree