| 摘要 |
The risk of an extreme gust of wind hitting a wind turbine is estimated by gathering data from one or more sensors for use as training data. This data is acquired over a period of time and is converted in to a feature vector for a given time period by a statistical measure. A number of feature zones are formed, each zone relating to a different estimate of risk with each feature vector being assigned to a risk category. The risk category is defined with reference to the value of one or more chosen turbine parameters at the time the data was acquired. The feature zones are formed from from a measure of distance such as the mean and covariance of feature vectors from within a given category. Live data is processed by measuring the mahalonobis distance from the feature vector of the live data to the centre of each zone and the risk of an extreme gust is assessed as that of the feature zone to which the mahalonobis distance is lowest. |
