Applying Swarm Intelligence and Bayesian Inference for Wind Turbine SCADA-Based Condition Monitoring and Prognostics

Xiang Ye and Lisa Osadciw
Submission Type: 
Full Paper
phmc_14_013.pdf606.82 KBAugust 30, 2014 - 1:53am

High operation and maintenance costs for wind turbines reduce their overall cost effectiveness. One of the biggest drivers of maintenance cost is unscheduled maintenance due to unexpected failures. Continuous monitoring of wind turbine health using automated failure detection algorithms can improve turbine reliability and reduce maintenance costs by detecting failures before they reach a catastrophic stage and by eliminating unnecessary scheduled maintenance. Furthermore, the prediction of under-performing or faulty turbines helps prevent these undesired working conditions and allows the operators to develop maintenance plans with prioritized tasks. This leads to shorter down-times and less revenue losses. Therefore, diagnosis and prognosis of potential faults is crucial to maintain and improve the efficiency of the wind energy system.

A SCADA-based condition monitoring and prognostics system uses data already collected at the wind turbine controller. It is a cost-effective way to monitor wind turbines for early warning of failures and performance issues. In this paper, we develop three tests on power curve, rotor speed curve and pitch angle curve of individual turbine. To monitor the turbine performance better in daily base, it is critical to recognize different patterns of turbine health condition by fusing all the test results. We apply particle swarm optimization algorithm to determine the fusion rules more objectively and optimally. This novel approach gains a qualitative understanding of turbine health condition to detect faults at an early stage, and also provides explanations on what has happened for detailed diagnostics.

As monitoring daily turbine health condition, we design a data-driven Bayesian inference approach to predict turbine potential failures with a percentage certainty by tracking the abnormal variations. We assume that the faulty turbine follows specific degradation patterns due to differential failure before and after it happens. Based on turbine degradation, the Bayesian network is able to tell the wind farm operators what type of failures is happening to this turbine, and tell the time when it is probably going to shut down. Also, the test results have verified the effectiveness of our approach.

Publication Year: 
Publication Volume: 
Publication Control Number: 
Page Count: 
Submission Keywords: 
wind energy
Fault Diagnosis and Prognosis
data-driven method
Bayesian inference
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Industrial applications
Submitted by: 

follow us

PHM Society on Facebook Follow PHM Society on Twitter PHM Society on LinkedIn PHM Society RSS News Feed