### Abstract

Language | English |
---|---|

Publication status | Published - 4 Feb 2013 |

Event | EWEA Annual Wind Energy Event 2013 - Vienna, Austria Duration: 4 Feb 2013 → 7 Feb 2013 |

### Conference

Conference | EWEA Annual Wind Energy Event 2013 |
---|---|

Country | Austria |

City | Vienna |

Period | 4/02/13 → 7/02/13 |

### Fingerprint

### Keywords

- copula statistics
- anomaly detection
- wind
- modelling

### Cite this

*Wind turbine operation anomaly detection using copula statistics*. Paper presented at EWEA Annual Wind Energy Event 2013, Vienna, Austria.

}

**Wind turbine operation anomaly detection using copula statistics.** / Stephen, Bruce; Gill, Simon; Galloway, Stuart; Wang, Yue; McMillan, David; Infield, David.

Research output: Contribution to conference › Paper

TY - CONF

T1 - Wind turbine operation anomaly detection using copula statistics

AU - Stephen, Bruce

AU - Gill, Simon

AU - Galloway, Stuart

AU - Wang, Yue

AU - McMillan, David

AU - Infield, David

PY - 2013/2/4

Y1 - 2013/2/4

N2 - The conventional means of assessing the performance of a wind turbine is through consideration of its power curve which provides the relationship between its active power outputs and a measured wind speed. An alternative representation can be obtained via the probability density estimation of the joint probability of wind speed and active power. A probabilistic model has several capabilities that could not be achieved with a simple curve fitting analysis, such as anomaly detection, which in practice may be related to the presence of an incipient problem; and model comparison which could allow the monitoring of the model evolution both over time and with respect to those of other turbines in a fleet. However, the probability density takes on a complex form which most probability distributions are incapable of representing. This paper illustrates the shortcomings of orthodox statistical approaches to probabilistic modelling of the power curve and shows how Copula statistics, traditionally used in finance models, offer a way of describing how to relate variables with a complex dependency structure in wind turbine condition monitoring.

AB - The conventional means of assessing the performance of a wind turbine is through consideration of its power curve which provides the relationship between its active power outputs and a measured wind speed. An alternative representation can be obtained via the probability density estimation of the joint probability of wind speed and active power. A probabilistic model has several capabilities that could not be achieved with a simple curve fitting analysis, such as anomaly detection, which in practice may be related to the presence of an incipient problem; and model comparison which could allow the monitoring of the model evolution both over time and with respect to those of other turbines in a fleet. However, the probability density takes on a complex form which most probability distributions are incapable of representing. This paper illustrates the shortcomings of orthodox statistical approaches to probabilistic modelling of the power curve and shows how Copula statistics, traditionally used in finance models, offer a way of describing how to relate variables with a complex dependency structure in wind turbine condition monitoring.

KW - copula statistics

KW - anomaly detection

KW - wind

KW - modelling

UR - http://www.ewea.org/annual2013/

M3 - Paper

ER -