On Wilcoxon rank sum test for condition monitoring and fault detection of wind turbines
Phong B. Dao
Abstract
In this study, the Wilcoxon rank sum test, a nonparametric statistical test method in the field of statistics, has been exploited for the operation state monitoring and automated fault detection of wind turbines. A five-step computation procedure has been developed in which the fault detection process is relied on statistical hypothesis tests. The null hypothesis is defined as that the wind turbine is operating in the normal condition without fault. If the null hypothesis is rejected in favour of the alternative, this indicates the occurrence of a fault in the wind turbine. The detection of a fault is indicated by an abrupt change from 0 to 1 in the test decision. The method can monitor one or several key process parameters of the wind turbine simultaneously. The monitoring process for each process parameter uses only its own data, thus, is independent with the monitoring process for other parameters. Hence, we do not need to handle the correlation between the selected parameters. Two wind turbine SCADA data sets, including three fault events, were used as case studies for testing the developed method. In both cases, various process parameters were analysed. The results show that the method can effectively monitor the operation state of wind turbines and reliably detect the anomaly (or fault) several days before its occurrence.