On wind Turbine failure detection from measurements of phase currents: a permutation entropy approach

TL;DR

This paper explores using permutation entropy to detect faults in wind turbines by analyzing phase current signals, showing that entropy measures can distinguish between healthy and faulty turbines.

Contribution

It introduces permutation entropy as a novel complexity measure for wind turbine fault detection, comparing it with traditional spectral analysis methods.

Findings

Permutation entropy varies significantly between healthy and faulty turbines.

Spectral analysis did not reveal differences in the current signals.

Higher entropy correlates with turbine failure in the studied data.

Abstract

This article presents the applicability of Permutation Entropy based complexity measure of a time series for detection of fault in wind turbines. A set of electrical data from one faulty and one healthy wind turbine were analysed using traditional FastFourier analysis in addition to Permutation Entropy analysis to compare the complexity index of phase currents of the two turbines over time. The 4 seconds length data set did not reveal any low frequency in the spectra of currents, neither did they show any meaningful differences of spectrum between the two turbine currents. Permutation Entropy analysis of the current waveforms of same phases for the two turbines are found to have different complexity values over time, one of them being clearly higher than the other. The work of Yan et. al. in has found that higher entropy values related to thepresence of failure in rotary machines in his study. Following this track, further efforts will be put into relating the entropy difference found in our study to possible presence of failure in one of the wind energy conversion systems.