Open AccessSupervisory control and data acquisition data‐based non‐linear state estimation technique for wind turbine gearbox condition monitoring
Author(s) -
Wang Yue,
Infield David
Publication year - 2013
Publication title -
iet renewable power generation
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.005
H-Index - 76
ISSN - 1752-1424
DOI - 10.1049/iet-rpg.2012.0215
Subject(s) - turbine , offshore wind power , wind power , computer science , condition monitoring , scada , fault (geology) , data acquisition , wind speed , range (aeronautics) , fault detection and isolation , reliability engineering , engineering , artificial intelligence , meteorology , mechanical engineering , physics , electrical engineering , aerospace engineering , seismology , actuator , geology , operating system
Offshore wind energy is catching increasing worldwide interest. However, access and maintenance offshore can be difficult and will be more costly than onshore, and hence, availability is correspondingly lower. As a result, there is a growing interest in wind turbine condition monitoring with condition‐based rather than responsive and scheduled maintenance. A non‐linear state estimation technique (NSET) model is presented here to model a healthy wind turbine gearbox using stored historical data. These data capture the inter‐relationship between the model input and output parameters. The state vectors comprising the data should cover as much as turbine operational range, including the extreme conditions in order to obtain an accurate model performance. A model so constructed can be applied to assess the operational data. Welch's t ‐test is employed in the fault detection algorithm, together with suitable time series filtering, to identify incipient anomalies in the turbine gearbox before they develop into catastrophic faults. Two case studies based on 10‐minute supervisory control and data acquisition data from a commercial wind farm are presented to demonstrate the model's effectiveness. Comparison is made with neural network modelling, and the NSET approach is demonstrated to be superior.