Wind turbine fault prognosis using SCADA measurements, pre-fault labeling, and KNN classifiers robust against data imbalance

In this article, an effective data-driven fault prognosis scheme is proposed using the KNN and its ensemble classifier. The operational dataset is labeled using the information contained in the status and warning datasets, and the labeled data, after eliminating invalid data, feature selection, and standardization is used for training and validation of the fault prognosis model. Using pre-fault labeling, fault prognosis is simplified to a fault diagnosis problem. The proposed method has provided good performance in fault prognosis compared to previous reports on this dataset. This is because, data imbalance, which is common in real data sets, does not affect the proposed method's performance, hence there is no need for data balancing methods in this algorithm and the performance is not deteriorated by false alarms. Also, the algorithm can determine labels of new observations online and only with their current feature values, without any requirement for historical data.