Sensorless effective wind speed estimation method based on unknown input disturbance observer and extreme learning machine

Abstract

Precise estimation of effective wind speed plays an important role in the advanced controls aiming at maximizing wind power extraction and reducing loads on turbine components. This paper proposes a sensorless effective wind speed estimation algorithm based on the unknown input disturbance observer and the extreme learning machine for the variable-speed wind turbine. First, aerodynamic torque is accurately estimated through an unknown input disturbance observer where the rotor speed is the measured output of the wind turbine drive train system. Then, the aerodynamic characteristics of the wind turbine are approximated by an extreme learning machine model based nonlinear input-output mapping. Last, effective wind speed is estimated based on the extreme learning machine model, using the previously estimated aerodynamic torque by the unknown input disturbance observer, together with the measured rotor speed and pitch angle. The proposed algorithm is validated by simulation studies on a 1.5 MW variable-speed wind turbine system. To evaluate the performance of the proposed algorithm, a detailed comparison with the Kalman filter-based method has been made. Comparison results clearly demonstrate that effective wind speed estimated by the proposed method is more accurate than that by the Kalman filter-based method and that the computational efficiency is higher.