Increasing application of power electronics devices (inverter-based generation and storage, HVDC lines, FACTS devices), facilitated by growing renewable energy integration, has also changed the ac grid modeling features. Namely, the incorporation of fast-acting control devices can lead to possible detrimental interactions that appear at much higher frequencies compared to synchronous generator-based power systems. In this regard, the phasor-based ac grid model is no longer adequate, putting forward the ac line models with extended validity in the frequency range. However, modeling all the lines in a system in high-frequency resolution requires high computational power and is often not necessary either. To overcome this challenge, this paper proposes the methodology for model order reduction (MOR) of an ac grid that enables a systematic identification of more significant lines in the system from the stability perspective and the necessary level of their model complexity. The methodology is established on subsystem participation calculation and the analysis of interaction modes between power electronics devices and a particular line. Thereat, the vector fitting algorithm is employed to obtain the high bandwidth state-space compatible ac line models. The functionalities and effectiveness of the proposed MOR methodology are demonstrated on a 9-bus test system.