Small-Scale Wind Turbine Recurrence and Cost Modeling as a Function of Operational Covariates from Supervisory Control and Data Acquisition Systems

Michael S. Czahor and William Q. Meeker
Submission Type: 
Full Paper
AttachmentSizeTimestamp
phmc_17_062.pdf779.66 KBAugust 17, 2017 - 7:51pm

Small-scale wind turbine (SWT) installations saw a dramatic increase between 2008 and 2012. Recently, the trend within industry has shifted towards installing larger wind turbines, leaving little attention for installed SWT reliability. Unfortunately, multiple downtime events raise concerns about the reliability and availability of the large number of installed SWTs. SWTs are repairable systems that can return to an operational state after a downtime or repair event. When a SWT experiences multiple events over time, these are known as recurrent events. The reliability of SWTs is examined in this paper using data from 21 individual 100 kW wind turbines. SWTs periodically record dynamic covariate data in the form of a vector time series using supervisory control and data acquisition (SCADA) systems. One type of event experienced by SWTs is known as a “service event,” which is a time when a SWT is put into service mode for a repair or false alarm. Due to the proprietary nature of the data used in this paper, different kinds of service events are combined, even though different failure modes and event types exist. We explore recurring service events and the associated cost of each “service event” and propose methodologies to link dynamic covariate data to downtime costs to assist in quantifying the variation of downtime across wind turbines. Data used in this work was provided from a power systems company in the United States. We outline a nonhomogenous Poisson process (NHPP) model with a Bayesian hierarchical power law structure for the count process and an autoregressive time series use rate model with a Bayesian framework to describe posterior parameter distributions. Using the posterior results, we develop a conditional and unconditional method to predict downtime mean cumulative functions (MCFs) for wind turbines.

Publication Year: 
2017
Publication Volume: 
8
Publication Control Number: 
062
Page Count: 
9
Submission Keywords: 
Bayesian Hierarchical
Prediction Intervals
Poisson process
wind energy
Submission Topic Areas: 
Data-driven methods for fault detection, diagnosis, and prognosis
Submitted by: 
  
 
 
 

follow us

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