This paper proposes an artificial intelligence-based approach to address the contribution of inverter-based resources (IBRs) in the system frequency response (SFR) model. IBRs are assumed to be aggregated within a dynamic virtual power plant (DVPP), a new entity that can be summoned by the system operator to provide ancillary services. Current SFR models in the presence of IBRs are computationally expensive and neglect the synchronization dynamics between conventional synchronous generators (SGs) and IBRs. In conventional SFR models, the interactions between SGs were formulated through a unique synchronization coefficient. However, in modern power systems, the interaction between SGs and IBRs within DVPPs should be represented by a 2 × 2 synchronization coefficients (SCs) matrix. Hence, an attempt is made to deal with the calculation of the SCs matrix using optimization algorithms and a neural network. More precisely, optimization algorithms are employed to calculate the SCs at different prespecified operating points, which are then fed into a neural network for training purposes, enabling online SFR derivation. The proposed approach is characterized by its low computational burden and complexity, which are critical for accommodating different levels of DVPP penetration.
