Back to the programme printer.gif Print




Delegates are invited to meet and discuss with the poster presenters in this topic directly after the session 'Advanced operation & maintenance' taking place on Thursday, 13 March 2014 at 11:15-12:45. The meet-the-authors will take place in the poster area.

Peter Tavner Durham University, United Kingdom
Co-authors:
Peter Tavner (1) F P Donatella Zappala (1) Chris Crabtree (1) Damian Vilchis-Rodriguez (2) Sinisa Djurovic (2)
(1) Durham University, Cambridge, United Kingdom (2) Manchester University, Manchester, United Kingdom

Printer friendly version: printer.gif Print

Presenter's biography

Biographies are supplied directly by presenters at EWEA 2014 and are published here unedited

Donatella Zappala will be the presenter of this paper. She graduated from Universita degli Studi, Perugia, in Environmental Engineering and then graduated from Durham Universiyt with an MSc in Renewable Energy. She is currently a final year PhD student with Peter Tavner, who is an Emeritus Professor of Durham Uniersity.

Abstract

Advanced algorithms for automatic wind turbine generator fault detection and diagnosis

Introduction

Previous wind turbine condition monitoring identified signals and methods for wind turbine fault detection. This paper concentrates on raising sensitivity so that high reliability is achieved. This is done by adopting techniques of earlier papers but aggregating them in a Side Band Power Factor algorithm that can be applied remotely and automatically to important WT drive train electrical and vibration signals. The process has already been demonstrated on a gearbox but will be extended here to consider generator electrical and vibration signals. The value of the algorithm and its detection sensitivity will be demonstrated by Test Rig results..

Approach

Previous wind turbine (WT) condition monitoring work by the authors concentrated on developing automatic algorithms for improving detection sensitivity on disparate parts of the WT drive train and proving their efficacy on a Test Rig.
The aim of automatic fault detection with improved sensitivity was based upon a need, identified from offshore wind farm operators [1], to determine offshore WT condition remotely, so that the costly mobilisation of diagnosis specialists could be minimised only to serious, repairable WT faults.
The authors have been involved in developing the following techniques:
• WT generator electrical fault detection, through analysis of fault frequency components or side-bands in electrical current or power spectra[2];
• Enhancing WT generator electrical and mechanical fault detection sensitivity by variable speed tracking of fault frequency side-bands in electrical and vibration spectra [3];
• Extending WT generator fault detection to consider rolling element bearing faults [4];
• Enhancing WT gearbox fault detection sensitivity by collating fault frequency side-bands in vibration spectra, using a Side Band Power factor (SBPF) algorithm [5].
In each case the efficacy of the methods was tested on the Durham University Wind Turbine Condition Monitoring Test Rig (WTCMTR) or a similar Test Rig at Manchester University.
In [5] that process was also extended to results from a full- WT gearbox, through the assistance of the National Renewable Energy laboratory (NREL) in the USA.
In the case of the WTCMTR it is possible to process the results in a commercial WT CMS, the SKF WindCon, further demonstrating the practicability of what was being proposed.
This paper will present combining and extending these methods, using SKF WindCon Observer, to produce higher detection sensitivities for WT variable speed generator electrical and mechanical faults, the algorithms will again be tested on the Durham WTCMTR and, if possible, on field results from a full-size WT generator.


Main body of abstract

Electrical Signals
An analysis of recorded generator current signals from the WTCMTR, using WindCon Observer, has shown that at different generator speeds, in the case of an unbalanced generator rotor fault there are clear amplitude increases of the 2sf upper sidebands of the supply frequency1st & 3rd harmonics. There is also clear dependence of the fault amplitude on the WT load, confirming what was shown for gearbox vibration signals in [2].
Similar results are expected in the power signals, down-shifted by the fundamental frequency close to DC.
The intent is to use these 2 side-band frequencies as a generator rotor unbalance fault indicator, using an SBPF algorithm similar to the gearbox work [5] and the work of [3] to track and automate fault detection. The analysis is still in progress and will be fully developed and tested on datasets already collected from the WTCMTR for the final Conference paper.
Current generator current simulations from Manchester have also shown that sidebands around the 1st & 3rd harmonics of the supply frequency are also related to the rotor unbalance. They only appear if the supply voltage unbalance is considered in the simulation. These will be useful rotor unbalance detection indicators under practical WT operating conditions when supply unbalance is unavoidable and these will be added into the planned SBPF algorithm.
Vibration Signals
The proposition is that the generator rotor resistance unbalance results in electromagnetic torque oscillations that induce mechanical vibration at the same frequencies in the machine frame. Given the close relationship between torque and power the frequencies that appear in the torque spectrum, as consequence of rotor resistance unbalance, will be present in the power and current signals. Therefore, the frequencies resulting from the rotor unbalance will be present in the power signal, as referred to in [2], and in the generator vibration signals [4].
An analysis of recorded generator accelerometer data from the WTCMTR, using WindCon Observer, has shown an amplitude increase, although not strong, of the 2sf upper sideband of the 2nd harmonic of the supply, when going from healthy to faulty conditions.
Manchester has run rotor unbalance tests, which allow a comparison between the two Test Rig generators’ vibration signatures under unbalanced rotor operation to confirm the proposition.
This work will be developed to provide an SBPF algorithm with a high sensitivity to detect rotor unbalance faults in the WT generator.
Combining Electrical & Vibration Signals
Finally the paper will combine the results of electrical and vibration signals, in the manner described above, for monitoring generator faults using the WindCon Observer software and the tracking algorithm IDFT developed in [3]. The results show an increased detection sensitivity compared to previous work and the potential to provide early detection of generator faults by application of the automatic SBPF & IDFT algorithm to diagnose faults on remote WTs before attendance at the WT.
Development of the Integrated Algorithm
Once the algorithm has been devised further tests will be done on the WTCMTR to demonstrate its effectiveness for a variety of loads and fault conditions and the results will be reported in the Conference paper.


Conclusion

This paper has shown that when a generator rotor has an unbalanced winding or a bearing fault that current, power and vibration signal spectra all show identifiable fault frequency sidebands, which can be tracked as the WT rotor speed varies.
A high fault detection sensitivity algorithm can be configured to track these side-bands using SBPF, significantly raising the detection sensitivity of the signals and improving the reliability of detection.
The results have also shown the benefit of tracking the SBPF using an IDFT algorithm as the WT power and speed varies.
These benefits have been demonstrated on Test Rigs under healthy and faulty conditions and the detection has been programmed in a commercial CMS, the SKF WindCon, using the Observer software provided.
Such algorithms could be deployed for large offshore wind farms to reliably identify generator rotor winding and bearing problems, known to be a significant cause of down-time.
The results of the paper will demonstrate this performance and show the ease with which such methods can be deployed in the practical environment.
It is hoped that field measurements from real WTs can be made available to demonstrate the application of these methods to real data beyond that of the authors’ Test Rigs.
The major learning outcomes of the paper are:
• To improve condition monitoring reliability using field fitted equipment;
• To devise automatic techniques that reduce the work-load on Wind Farm Operators;
• To develop more comprehensive techniques to monitor large offshore wind farms;
• To demonstrate that cheap and effective continuous condition monitoring is feasible on large offshore wind farms.



Learning objectives
To improve condition monitoring reliability using field fitted equipment;
To devise automatic techniques that reduce the work-load on Wind Farm Operators;
To develop more comprehensive techniques to monitor large offshore wind farms;
To demonstrate that cheap and effective continuous condition monitoring is feasible on large offshore wind farms.



References
[1] Feng, Y, Tavner, P.J, Long, H, Early experiences with UK Round 1 offshore wind farms, Proc Institution of Civil Engineers: Energy. 163(4), 2010: 167-181.
[2] Djurovic, S, Crabtree, C.J, Tavner, P.J, Smith, A.C, Condition monitoring of wind turbine induction generators with rotor electrical asymmetry, IET Renewable Power Generation, 6(4), 2012: 207-216.
[3] Yang, W, Tavner, P.J, Crabtree, C.J, Wilkinson, M.R, Cost-effective condition monitoring for wind turbines, IEEE Transactions on Industrial Electronics, 57(1), January 2010: 263-271.
[4] Vilchis-Rodriguez, D.S, Djurović, S, Smith, A.C, Wound rotor induction generator bearing fault modelling and detection using stator current analysis, IET Renewable Power Generation 7(4), 2013: 330-340.
[5] Zappalà, D, Tavner, P.J, Crabtree, C.J, Sheng, S, Sideband algorithm for automatic wind turbine gearbox fault detection and diagnosis, Proc Scientific Track, European Wind Energy Association Conference, EWEA2013, Vienna 2013.