Answers are grounded in this paper and cite their sources.
Converter-interfaced energy storage now carries a growing share of primary frequency regulation, but its two control modes fail in opposite ways: grid-following (GFL) units can be rated large yet act through a phase-locked loop and a measurement delay, while grid-forming (GFM) units act at once but are small and tightly current limited. This paper proposes a predictive complementary split with a robust frequency-security guard (PCS-RG). When grid connected, the GFM unit supplies virtual inertia and a wash-out droop while the GFL unit applies an event-triggered, front-loaded droop to a frequency predicted from the rate of change of frequency (RoCoF), which offsets the GFL delay without forecasting the nadir. After islanding the GFM unit follows an energy-aware trajectory of inertia, transient droop and ramped steady droop, and a guard identifies the power imbalance and the governor droop online, raising the droop only when the steady-state limit is at risk; if synchronous machines remain, the GFL unit re-locks and carries that droop. On a three-area average-system-frequency model the scheme lowers the nadir by 9.6% at unchanged state of charge, keeps it 33% lower at 300 to 400 ms of delay, cuts islanded RoCoF by 30% and removes every steady-state violation in 120 randomised islands, where a deviation-scheduled benchmark violates in 48% of cases. A validation on the IEEE 10-machine 39-bus system with machine dynamics and explicit converter current limits gives the benchmark nadir with 41% less GFM peak power after a generator trip and, in a 15-bus island with a 548 MW deficit, a 0.161 Hz nadir and a 0.068 Hz settling deviation where the benchmark saturates its converter, exhausts it after 15 s and settles outside a 0.12 Hz limit.
No takes yet. Share an insight, caveat, or question.
Muzeeb Khan Patan (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: