Szczegóły publikacji
Opis bibliograficzny
On cointegration analysis for condition monitoring and fault detection of wind turbines using SCADA data / Phong B. DAO // Energies [Dokument elektroniczny]. — Czasopismo elektroniczne ; ISSN 1996-1073. — 2023 — vol. 16 iss. 5 art. no. 2352, s. 1–17. — Wymagania systemowe: Adobe Reader. — Bibliogr. s. 16–17, Abstr. — Publikacja dostępna online od: 2023-03-01
Autor
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 145530 |
|---|---|
| Data dodania do BaDAP | 2023-03-14 |
| Tekst źródłowy | URL |
| DOI | 10.3390/en16052352 |
| Rok publikacji | 2023 |
| Typ publikacji | artykuł w czasopiśmie |
| Otwarty dostęp |  |
| Creative Commons | |
| Czasopismo/seria | Energies |
Abstract
Cointegration theory has been recently proposed for condition monitoring and fault detection of wind turbines. However, the existing cointegration-based methods and results presented in the literature are limited and not encouraging enough for the broader deployment of the technique. To close this research gap, this paper presents a new investigation on cointegration for wind turbine monitoring using a four-year SCADA data set acquired from a commercial wind turbine. A gearbox fault is used as a testing case to validate the analysis. A cointegration-based wind turbine monitoring model is established using five process parameters, including the wind speed, generator speed, generator temperature, gearbox temperature, and generated power. Two different sets of SCADA data were used to train the cointegration-based model and calculate the normalized cointegrating vectors. The first training data set involves 12,000 samples recorded before the occurrence of the gearbox fault, whereas the second one includes 6000 samples acquired after the fault occurrence. Cointegration residuals—obtained from projecting the testing data (2000 samples including the gearbox fault event) on the normalized cointegrating vectors—are used in control charts for operational state monitoring and automated fault detection. The results demonstrate that regardless of which training data set was used, the cointegration residuals can effectively monitor the wind turbine and reliably detect the fault at the early stage. Interestingly, despite using different training data sets, the cointegration analysis creates two residuals which are almost identical in their shapes and trends. In addition, the gearbox fault can be detected by these two residuals at the same moment. These interesting findings have never been reported in the literature.
