An improved-accuracy method for fatigue load analysis of wind turbine gearbox based on SCADA

Abstract

In spite of their increasing popularity, managing the use of wind turbines has been exceptionally challenging. Through computational prognostics, Sentient Science determined that current operating lifetime for a large number of turbines is only between five to thirteen years. Initial estimates indicate that savings of $150,000 per turbine per gearbox replacement can be achieved using physics-based long-term prognostics, leading to a substantial return of investment for wind farm operators. However, long-term prognostics require a precise determination of the loads in all six degrees of freedom occurred on the drive-train. One of these loads—torque—can be directly estimated in situ from the historical data provided by the Supervisory Control and Data Acquisition (SCADA) system. In many cases, the historical data only provides 10-min statistical values, and a common practice of reliability analysts is the calculation of torque using only 10-min averages. Disregarding the load fluctuation within 10-min intervals of recorded SCADA introduces a loss of accuracy in the resulting torque histogram that is indeed meaningful for an accurate life prognostic. This paper introduces a novel improved-accuracy method for calculation of torque histograms based on SCADA. Using 10-min distributions of power output and rotor speed, this method is able to successfully reconstruct the distribution of instantaneous torque in between 10-min intervals of recorded SCADA. The method predicts a high-torque region more dispersed that the current method used in the industry, which introduces substantially different results when used in life prognostics. Using this method in the lifing of a GE 1.5 SLE wind turbine, it is shown that the error in predicted L50 is reduced by 10.1%.