Online Model-Based Redesign of Experiments for Parameter Estimation Applied to Closed-loop Controller Tuning

We present an approach to closed-loop online model-based redesign of experiments for system identification. Special attention is given to the compliance with safety restrictions and operating requirements during online experiments. For doing so, we propose the integration of a controller into the system identification algorithm. To avoid numerical problems regarding ill-conditioned matrices an algorithm for local parameter identifiability analysis is used. In order to demonstrate the benefits of our approach, the proposed procedure is validated in a real case study. Additionally, a PI-controller is tuned for the identified system. Moreover, the accuracy of the system parameters estimated by the proposed strategy was compared with results for a conventional open-loop step response technique.