Monitoring Wind Turbine Vibration Based on SCADA Data

作者：Zhang Zijun^*; Kusiak Andrew

来源：Journal of Solar Energy Engineering-Transactions of the ASME, 2012, 134(2): 021004.

DOI：10.1115/1.4005753

• turbine vibration
• monitoring
• control chart
• k-means clustering
• drivetrain acceleration
• tower acceleration
• data-mining
• neural networks ensemble

Three models for detecting abnormalities of wind turbine vibrations reflected in time domain are discussed. The models were derived from the supervisory control and data acquisition (SCADA) data collected at various wind turbines. The vibration of a wind turbine is characterized by two parameters, i.e., drivetrain and tower acceleration. An unsupervised data-mining algorithm, the k-means clustering algorithm, was applied to develop the first monitoring model. The other two monitoring models for detecting abnormal values of drivetrain and tower acceleration were developed by using the concept of a control chart. SCADA vibration data sampled at 10 s intervals reflects normal and faulty status of wind turbines. The performance of the three monitoring models for detecting abnormalities of wind turbines reflected in vibration data of time domain was validated with the SCADA industrial data. [DOI: 10.1115/1.4005753]

• 出版日期2012-5