Towards a More Robust Understanding of the Uncertainty of Wind Farm Reliability

Abstract

Understanding wind farm reliability from various data sources is highly complex because the boundary conditions for the data are often undocumented and impact the outcome of aggregation significantly. Sandia National Laboratories has been investigating the reliability of wind farms through the Continuous Reliability Enhancement Wind (CREW) project since 2007 through the use of Supervisory Control and Data Acquisition (SCADA) data from multiple wind farms in the fleet of the USA. However, data streaming from sample wind farms does not lead to better understanding as it is merely a generic status of those samples. Economic type benchmark studies are used in the industry, but these do not yield much technical understanding and give only a managerial cost perspective. Further, it is evident that there are many situations in which average benchmark data cannot be presented in a meaningful way due to discrete events, especially when the data is only based on smaller samples relative to the probability of the events and the sample size. The discrete events and insufficient descriptive tagging contribute significantly to the uncertainty of a fleet average and may even impair the way we communicate reliability. These aspects will be discussed. It is speculated that some aspects of reliability can be understood better through SCADA data-mining techniques and considering the real operating environment, as, it will be shown that there is no particular reason that two identical wind turbines in the same wind farm should have identical reliability performance. The operation and the actual environmental impact on the turbine level are major parameters in determining the remaining useful life. Methods to normalize historical data for future predictions need to be developed, both for discrete events and for general operational conditions.