Bayesian Network for Wind Turbine Fault Diagnosis

Chen, Bindi; Tavner, P.J.; Feng, Y.; Song, W.W.; Qiu, Y.N.

Abstract

Wind turbine reliability studies have become more important because good wind turbine reliability with predictable turbine maintenance schedule will reduce the cost of energy and determine the success of a wind farm project. Previous research on wind turbine SCADA system has made progress in this respect. However, SCADA data volume is usually too huge and alarm information is too unclear to indicate failure root causes. In addition, SCADA signals and alarms are not currently interpreted as a whole. This highlights the need for more intelligent methods which can use existing SCADA data to automatically provide accurate WT failure diagnosis. This paper presents a new approach, based on Bayesian Network, to describe the relationship between wind turbine failure root causes and symptoms. The Bayesian Network model was derived from an existing probability-based analysis method – the Venn diagram, and based upon 26 months of historical SCADA data. The Bayesian Network reasoning results have shown that the Bayesian Network is a valuable tool for WT fault diagnosis and has great potential to rationalise failure root causes.

Citation

Chen, B., Tavner, P., Feng, Y., Song, W., & Qiu, Y. (2012). Bayesian Network for Wind Turbine Fault Diagnosis.

• Presentation Conference Type: Conference Paper (Published)
• Conference Name: EWEA 2012
• Start Date: Apr 16, 2012
• End Date: Apr 19, 2012
• Publication Date: Apr 17, 2012
• Deposit Date: Jul 10, 2013
• Publicly Available Date: Jul 12, 2013
• Keywords: Wind Turbine, Bayesian Network, SCADA, Fault Diagnosis.
• Public URL: https://durham-repository.worktribe.com/output/1156248
• Publisher URL: http://proceedings.ewea.org/annual2012/proceedings/#