This paper presents an approach to obtain reduced-order models for power networks involving power electronic converters (PEC) via the frequency-domain balanced realizations (FDBR) technique. PECs play an essential role in power processing and energy conversion in modern electrical networks, such as the interconnection of renewable generators, HVDC links, and active filters. Integration of PECs into dynamic equivalents needs model-order reduction (MOR) in both low- and high-frequency ranges to account for both slow and fast dynamics due to the network and switching natures. The objective of the FDBR technique is to obtain an internally balanced system, i.e., an equally controllable/observable system, that can be reduced according to its dominant dynamics within the limited frequency bandwidths. This allows accounting for specific band-limited phenomena, such as those generated within a power network caused by PECs, which is the focus of this paper. The results show that faster yet accurate simulations are achieved by reduced-order models through FDBR compared to their full-order counterparts.