Data-driven three-step optimal frequency scheme for alleviating grid frequency regulation in MT-HVDC linked grids

The increasing penetration of renewable energy sources and the deployment of multi-terminal high-voltage direct-current (MT-HVDC) systems complicate reliable grid frequency regulation. This paper proposes a data-driven three-step optimal frequency control scheme for MT-HVDC systems. The first step focuses on data-driven state-space modeling using observer/Kalman filter identification and the eigensystem realization algorithm. The data-driven approach constructs reduced-order small-signal models that accurately capture system dynamics using only input–output data, independent of first-principles descriptions. In the second step, an optimal linear quadratic Gaussian (LQG) control scheme is synthesized based on the identified model. The LQG control algorithm minimizes deviations in frequency and DC-link voltage via a quadratic optimal-control formulation. Concurrently, the embedded Kalman filter ensures robust state estimation under measurement noise and uncertainties. The third step implements an active power compensation (APC) algorithm to enhance resilience under abnormal operating conditions, such as inverter malfunctions and communication delays. The APC algorithm reallocates power references among healthy converters, preserving stability and damping transients. Extensive simulations validate the effectiveness of the proposed framework, demonstrating its ability to achieve accurate system modeling, robust frequency regulation, and fault-tolerant operation.