Multi-Timescale Reduced-Order Modeling of Direct-Drive Wind Turbines for On-Demand Simulation Across Diverse Grid-Integration Scenarios
Full-order models of direct-drive wind turbines capture microsecond switching transients through second-scale control dynamics but impose prohibitive computational burdens for large-scale or long-duration grid-integration studies. Existing reduced-order approaches—controlled current source equivalents, empirical timescale truncation, linearization with balanced truncation, singular perturbation decomposition, local model switching, and data-driven surrogates—lack a systematic framework for matching model complexity to the dominant response of a specific analysis task. This paper proposes an on-demand simulation framework that partitions turbine dynamics into six hierarchical timescales: switching, AC current, mid-frequency, DC voltage, rotor speed, and operational. Each level is mapped to its dominant physical components, critical state variables, and typical grid-study scenarios, establishing explicit correspondences between analysis objectives and required model fidelity. A reduction methodology combining timescale-dominant response identification with trajectory sensitivity analysis is developed to selectively retain fast dynamics within the target timescale while enforcing steady-state consistency constraints on slower variables. Case studies at the AC current and rotor speed timescales demonstrate that the reduced models reproduce fast current and voltage transients with approximately 5–6× simulation speedup while preserving steady-state accuracy, and accurately capture power transient evolution governed by rotor inertia and slow-variable regulation. The framework provides a structured pathway for constructing hierarchical, precision-controllable turbine models for large-scale wind farm simulation and multi-scenario grid-integration analysis.