Detection of wind turbine power performance abnormalities using eigenvalue analysis

Abstract

Condition monitoring of wind turbines is a field of continuous research and development as new turbine configurations enter into the market and new failure modes appear. Systems utilising well established techniques from the energy and industry sector, such as vibration analysis, are commercially available and functioning successfully in fixed speed and variable speed turbines. Power performance analysis is a method specifically applicable to wind turbines for the detection of power generation changes due to external factors, such as icing, internal factors, such as controller malfunction, or deliberate actions, such as power de-rating. In this paper, power performance analysis is performed by sliding a time-power window and calculating the two eigenvalues corresponding to the two dimensional wind speed-power generation distribution. The power is classified into five bins in order to achieve better resolution and thus identify the most probable root cause of the power deviation. An important aspect of the proposed technique is its independence of the power curve provided by the turbine manufacturer. It is shown that by detecting any changes of the two eigenvalues trends in the five power bins, power generation anomalies are consistently identified.