Data Driven, Additive Fault Detection for Wind Turbines

Abstract

This paper covers an approach for fault residuals construction, when carrying out the data driven techniques. The method is based on the system input and output data and is built on a difference between data of a “normal” and a “faulty” system behavior. The research evaluates the performance of the structures, based on the subspace identification technique, to highlight the acceptable usefulness of the resulting data driven approach. Following the same lines, it delineates the connection to input/output data and provides an approach to computation of the parameterized data matrices. The principle is applied directly in analysis of the data generated by the wind turbine model.