Comparative study of binning and gaussian process based rotor curves of a wind turbine for the purpose of condition monitoring

The wind turbines deteriorating performance cause failures, including catastrophic failures that lead to high operation and maintenance (O&M) cost. SCADA based continuous monitoring of wind turbines is a cost-effective approach and plays an essential role as turbine sizes increase, and they are placed in more remote locations, for example, offshore to improve the performance of wind turbines and reduces the O&M cost. Various studies suggest that the internal operation of the wind turbine depends on various variables, especially on rotor speed. The proper analysis of rotor speed can be useful for constructing effective SCADA based models for wind turbine condition monitoring. Gaussian Process (GP) is a nonparametric, stochastic process that's designed to solve regression and probabilistic classification problems. GP models powerful to solve nonlinear systems, however, its application are not much explored in wind turbine condition monitoring.This paper describes a wind turbine condition monitoring Gaussian Process technique that uses the rotor speed to derive a rotor curve for wind turbine condition monitoring. Developed GP model, then compared with the conventional approach based on binned rotor curves together with individual bin probability distributions to identify operational anomalies. The proposed techniques have been validated experimentally using SCADA data sets obtained from operational turbines. Finally, comparative analysis of these techniques described outlining the strength and weakness of individual models.