The expansion of renewable energies, onshore often in the form of distributed generation plants (DGs), is shifting power generation out of the transmission grids and into the distribution grids. As a result, in addition to generation itself, more and more ancillary services are required of DG, among other characteristics, to compensate for such functionalities of large power plants that are no longer needed. In sum, the operating resources and functions of the distribution grid system, consisting of the grid itself and the connected customers, are considerably increased, resulting in a significantly more complex non-linear dynamic behavior between the electrical input and output variables of the now active distribution grids than was previously the case. The development of time-synchronized phasor measurement units (PMU) allows the implementation
of wide area monitoring systems (WAMS) with the associated data and computer infrastructure. As a result, time-synchronous measurement data are increasingly available at higher and higher resolutions up to the accuracy required for mapping transient processes. This work is dedicated to the analysis of the behavior of active distribution networks, with respect to the determination of models or equivalents of these by means of measurement data recorded in operation from WAMS, as well as the analysis of the properties of common model types and identification methods with respect to their parameterizability by means of operational measurement data. In addition, this work provides a direct comparison between the most commonly used approaches in literature, black-box approaches based on artificial neural networks and grey-box approaches based on parametric models.
In the analysis of the behavior, the quasi-stationary nonlinearity between the input and output variables of the distribution network is proposed as a suitable evaluation measure of its complexity or the complexity of the identification problem of a distribution network. The discussion of the causes of this nonlinearity, depending on the behavior of DGs, allows to transfer the knowledge collected for individual concrete model networks to other distribution networks depending on the composition of loads and types of DG units. The tests of the equivalents are each embedded in the network calculation program within longer simulation runs in order to demonstrate the accuracy and stability of the equivalents under real conditions of simulation studies.