the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
Unsupervised anomaly detection of permanent magnet offshore wind generators through electrical and electromagnetic measurements
Abstract. This paper investigates fault detection in offshore wind permanent magnet synchronous generators (PMSG) for demagnetization and eccentricity faults (both static and dynamic) at various severity levels. The study utilizes a high-speed PMSG model, on the NREL 5-MW reference offshore wind turbine, and at the rated wind speed, to simulate healthy and faulty conditions. An unsupervised convolutional autoencoder (CAE) model, trained on simulated signals from the generator in its healthy state, serves for anomaly detection. The main aim of the paper is to evaluate the possibility of fault detection by means of high-resolution electrical and electromagnetic signals, given that the typically low-resolution standard measurements used in SCADA systems of wind turbines often impede the early detection of incipient failures. Signals analyzed include three-phase currents, induced shaft voltage, electromagnetic torque, and magnetic flux (airgap and stray) from different directions and positions. The performance of CAE models is compared across time and frequency domains. Results show that in the time domain, stator three-phase currents effectively detect faults. In the frequency domain, stray flux measurements, positioned at the top, bottom, and sides of outside the stator housing, demonstrate superior performance in fault detection and sensitivity to fault severity levels. Particularly, radial components of stray flux can successfully distinguish between eccentricity and demagnetization.
- Preprint
(11409 KB) - Metadata XML
- BibTeX
- EndNote
Status: open (until 11 May 2024)
-
RC1: 'Comment on wes-2024-26', Anonymous Referee #1, 24 Apr 2024
reply
The manuscript provides a method to detect demagnetization and eccentricity in a permanent magnet synchronous generator using an unsupervised convolution autoencoder model. The method uses measured signals to detect the fault. The manuscript effectively outlines its objective and hypotheses. However, it is recommended that the authors address the following comments to enhance the clarity of the manuscript:
- Provide a comprehensive literature survey on the fault detection method using machine learning and compare it with the proposed method.
- How the shaft voltage in Fig. 4 (d) is simulated? Is the impact of the converter excitation and the bearing current considered during simulation?
- How the eccentricity is simulated. Quantify the misalignment of the rotor relative to the stator?
- Fig. 13 (e-f) shows no change in the power spectral density of the phase current between the healthy condition and fault condition. Are the time-domain signals also same? If yes, then how it is used in section 4.2 to detect the fault condition?
- Does the accuracy of the fault detection is influenced by the sampling frequency of the signal?
Citation: https://doi.org/10.5194/wes-2024-26-RC1
Viewed
HTML | XML | Total | BibTeX | EndNote | |
---|---|---|---|---|---|
171 | 42 | 7 | 220 | 4 | 2 |
- HTML: 171
- PDF: 42
- XML: 7
- Total: 220
- BibTeX: 4
- EndNote: 2
Viewed (geographical distribution)
Country | # | Views | % |
---|
Total: | 0 |
HTML: | 0 |
PDF: | 0 |
XML: | 0 |
- 1