Identification of parametric models in the frequency-domain through the subspace framework under LMI constraints

Abstract

An extension of standard subspace-based algorithms to identify parametric systems from frequency response samples is presented. This algorithm uses arbitrary frequency spacing and allows frequency weighting. It offers the possibility to impose the model's poles location through LMI constraints. This technique is applied to a numerical example and to real industrial frequency-domain data originating from an openchannel flow for hydroelectricity production.

Data-driven modeling of power networks

“…The least-squares problem (20) is linear in the unknown variables A r , H r and B r . Hence, the minimization problem (20) can be transformed into solving the linear least-squares problem [28], [30] min X ∈R (r+ r 2 +r 2 +q)×r…”

“…Various methods have been used to construct a reduced model from data. While some approaches are based on the frequency-domain data (see, e.g., [1], [2], [11], [14], [16], [17], [20], [23]) the others use time-domain data (see, e.g., [12], [19], [21], [30], [32], [34]).…”