Fault detection and isolation in wind turbines using support vector machines and observers

In this work, the benchmark FAST that simulates a closed-loop three-bladed wind turbine is used for fault detection and isolation. Two methods were employed to isolate faults of different types at different locations: Support vector machines (SVM) and a Kalman-like observer. SVM could isolate most faults with the used data and characteristic vectors, except for high varying dynamics. In this case, the use of an observer, which is model-based, was found necessary.