Simple

Data driven approaches in fault detection of wind turbines

Dusan Krokavec
Department of Cybernetics and AI, FEI, Technical University of Kosice

Anna Filasova
Department of Cybernetics and AI, FEI, Technical University of Kosice

Abstract
The paper gives an approach to fault residuals construction performing the data driven technique. The method comes out from the system input and output representations and is constructed on a collection of measured data of normal and faulty system behaviour. The paper evaluates the performance of an FDI structure based on subspace identification technique to highlight the some usefulness of the data driven approaches. Proceeding along the same lines it is outlined the connection to input/output data and given an approach to identification of parameterised matrices and the fault detection logic. The principle is applied directly in analysis of the data generated by monitored system wind turbine model parameters.