Int J Performability Eng ›› 2013, Vol. 9 ›› Issue (3): 245-260.doi: 10.23940/ijpe.13.3.p245.mag

• Original articles • Previous Articles     Next Articles

Clustering Analysis to Improve the Reliability and Maintainability of Wind Turbines with Self-Organizing Map Neural Network

ZAFAR HAMEED and KESHENG WANG   

  1. Norwegian University of Science and Technology (NTNU), Trondheim, NORWAY

Abstract:

Reliability and maintainability of wind turbines are posing new challenges and issues due to advancements in new and sophisticated technologies. This has necessitated development of novel, efficient, and cost-effective strategies for enhancing the availability for power output and operational life. There are certain ways to achieve such objectives where every approach has certain pros and cons. One possible technique is to use the wind speed and power output data for exploring the behavioral similarities of different wind turbines. Based on the similarity measures, a group of turbines may appear together in the form of clusters. This is accomplished by working with the vast piles of data which are pre-processed by using statistical time domain features to provide input to a self-organizing map (SOM) neural network. Based on the clustering results, operational and maintenance strategies are planned for a group of wind turbines in contrast to doing the same work for individual ones. A case study is presented where it has been shown how the information obtained from the clustering analysis would be used for predicting the power output and then developing the optimal operational and maintenance strategies in an integrated manner.


Received on May 03, 2012, revised on September 03, 2012
References: 28