Blade-pitch dynamics in DFIG wind turbines: modal sensitivity, transient loading, and weak-grid feasibility

Abstract Small-signal and sub-synchronous stability studies of doubly-fed induction generator (DFIG) wind turbines almost universally treat the blade-pitch angle as a frozen parameter, assuming pitch actuation is too slow to interact with electrical and torsional modes—an assumption widely used but rarely quantified. Using a high-fidelity 24-state model of a 3.6 MW DFIG turbine that retains the fourth-order machine, two-mass drivetrain, rotor- and grid-side converter controls, a phase-locked loop (PLL), and a dynamic pitch controller, this paper compares the modal spectrum and the large-signal response with pitch dynamic versus frozen, across short-circuit ratios (SCR) from $$\\approx 43$$ to $$\\approx 3$$ and the pitch-active wind range. Pitch dynamics shift the damping of every electrical, network, converter, and torsional mode by less than $$1\\times 10^{-3}$$ , and the corresponding eigenvalue, modal frequency and mode-shape deviations remain below $${2\\times 10^{-3}}$$ (relative), $${4\\times 10^{-4}}\\,\\textrm{Hz}$$ and $${1.1}^\\circ $$ respectively, justifying frozen pitch specifically for modal and sub-synchronous resonance screening and for impedance-based screening, but not for large-disturbance, overload, voltage-recovery or thermal/mechanical-limit assessment. The sustained large-signal response, however, is strongly pitch dependent: for a finite-rate wind ramp to rated-plus conditions, freezing pitch leaves a sustained power overload of up to $$54\\%$$ of rated apparent power, held indefinitely, and a depressed weak-grid point-of-common-coupling (PCC) voltage that dynamic pitch removes; a voltage-dip ride-through swept over dip depth and duration confirms the dual—fast grid-side transients are pitch-invariant. The discriminant is disturbance duration relative to the pitch time constant, and a parametric study places the transition at $$T_d/T_\\beta \\approx {0.3}$$ . A weak-grid sweep further shows the feasibility limit of this configuration to be a voltage-loadability (PV-nose) limit near SCR $$\\approx $$ 2.3 rather than a damping limit, extendable to SCR $$\\approx $$ 1.7 by converter reactive support. Modeling refinements for a well-posed PLL equilibrium and a consistent network coupling are also documented, and the frozen-pitch reduction is shown to cut eigenvalue and sweep cost by 25– $${30}\\%$$ .