Early Fault Diagnosis Strategy for WT Main Bearings Based on SCADA Data and One-Class SVM

To reduce the levelized cost of wind energy, through reduction in operation and maintenance costs, it is imperative that turbine downtime reduced strategies based on condition monitoring. The standard approach toward this challenge vibration monitoring, which requires installation specific tailored sensors incur associated added costs. On other hand, life expectancy parks built during 1990s power boom dwindling, data-driven issued from already accessible supervisory control data acquisition (SCADA) an auspicious competitive solution because no additional are required. Note a major issue to provide fault diagnosis approaches only SCADA data, as these were not established with objective being used for monitoring but rather capacities. present study posits early strategy exclusively supports results real park 18 turbines. contributed methodology anomaly detection model one-class support vector machine classifier; is, semi-supervised trains decision function categorizes fresh similar or dissimilar training set. Therefore, healthy (normal operation) required train model, greatly expands possibility employing (because there need faulty past, normal needed). obtained show promising strategy.