Support Vector Machines for Fault Detection in Wind Turbines

Abstract

AbstractSupport Vector Machines (SVM) are used for fault detection and isolation in a variable speed horizontal-axis wind turbine composed of three blades and a full converter. The SVM approach is data based and is therefore robust to process knowledge. Moreover, it is based on structural risk minimization which enhances generalization and it allows accounting for process non linearity by using flexible Kernels. In this work, a radial basis function was used as Kernel. Different parts of the process were investigated including actuators, sensors and process faults. With duplicated sensors, we could detect sensor faults in blade pitch positions, generator and rotor speeds rapidly. Fixed value fault were detected in 2 sample periods and offset faults could be detected for Δβ ≥ 0.5° with a detection time that depends on the offset level. The converter torque fault (an actuator) could be detected within two sample periods. Faults in the actuators of the pitch systems could not be detected. Faults in the process concerning friction in the drive train could be detected only for very high offset (Δηdt ≥ 50%).