Identification of LTI models from concatenated data sets

Research output: Unpublished contribution to conferenceUnpublished paper


For some industrial applications, experimental data is available in the form of several data sets corresponding to the operation of the plant under the same conditions. An example of such an application is the condition monitoring of a wind turbine based on SCADA data. Here, one is interested in the identification of a turbine subsystem for a specific wind condition. However, long records of a given operating condition might be difficult to obtain. Hence, one needs to select multiple short data-records from the operational data to identify the system. In this case, identification approaches where missing data are treated as unknown parameters [1, 2] are not feasible due to the large amount of lost data. Then, the best option is to concatenate the data sets, and introduce additional parameters to handle the transient effects [3]. Our aim is to verify the consistency of the estimates when considering this last approach. To this end, we analyzed the estimator cost function and performed a Monte Carlo simulation to prove consistency. A previous work [4] showed that AR and ARX are consistent, this time we focus on the Output Error (OE) model structure.
Original languageEnglish
Publication statusPublished - Mar 2019
Event38th Benelux Meeting on Systems and Control - Center Parcs De Vossemeren, Lommel, Belgium
Duration: 19 Mar 201921 Mar 2019


Conference38th Benelux Meeting on Systems and Control
Internet address


Dive into the research topics of 'Identification of LTI models from concatenated data sets'. Together they form a unique fingerprint.

Cite this