A physics-based approach to wind turbine SCADA data analysis and power curve outlier explanation

Abstract

In a world where the energy demands only increase, new forms of renewable energy are critical. Amidst the need for new forms of energy, offshore wind energy is a growing industry with promising results. Although it is a promising solution, the industry faces several challenges related to maintenance. As a means to reduce the need for corrective maintenance and repair, preventive and predictive maintenance should be increased. One approach to achieve this is to utilise tools such as the wind turbine power curve. The curve shows the relationship between the wind speed and the wind turbine's power output.

The SCADA data set contains outliers which should be explained for a better understanding of the wind turbine operations as well as for fault prediction purposes. The thesis aims to use a physics-based approach to analyse SCADA data and to explain outliers in the power curve.

Outliers in the SCADA data set will present themselves in the wind turbine power curve. Often when the power curve is cleaned, the outliers are removed without an explanation. A physics-based approach to analyse the SCADA data, and thus the outliers in the power curve, is carried out by creating rules based on how the wind turbine is expected to operate. This requires an understanding of how the wind turbine works and essentially how the various components affect one another.

Using a physics-based approach to analyse the SCADA data and creating rules for expected operational behaviour to identify outliers, proved to be effective. The outliers were provided explanations. Although the majority of outliers were due to operational states and transitions, the two generator bearing failures as well as the failure in the generator were detected.