Hybrid mechanism-data-driven modeling for power system frequency response.

The integration of large-scale renewable energy into power systems has led to a significant reduction in equivalent inertia and primary frequency regulation capability, posing severe challenges to frequency stability. To accurately analyze the frequency dynamic response trajectory of power systems, a hybrid mechanism-data-driven modeling scheme is proposed. First, an extended system frequency response model that accounts for boiler thermodynamics is developed. Using this model, the time-varying characteristics of different parameters are analyzed, leading to the establishment of a hybrid modeling framework. Then, the auto-regressive moving average with exogenous inputs model is utilized to describe the input–output relationship between active power disturbance and system frequency dynamics, with parameters identified via the forgetting factor recursive least squares method. Moreover, the gated recurrent unit network integrated with an attention mechanism is employed to estimate the thermal state, thereby enabling the calculation of frequency regulation capability. Finally, case study results demonstrate that the proposed scheme can reduce the modeling error and improve the analysis accuracy of frequency dynamics, which verifies its effectiveness.