Real-Time Adaptive Virtual Inertia Control of Grid-Forming Inverters under Frequency Disturbances
Annotatsiya
Adaptive virtual inertia can make grid-forming (GFM) inverters both more effective and more economical, but practical adaptive laws must operate with measurement noise and latency, and against disturbances other than a single step. This paper proposes a real-time adaptive law in which a disturbance observer (DOB) estimates the current and the cumulative power imbalance from the swing equation, using the measured inverter power and a filtered RoCoF. The virtual inertia and damping are then set from closed-form RoCoF and frequency-deviation requirements, with peak holding and rate limits. Under 20 ms latency and 2 mHz noise, the law was compared with fixed, bang-bang and smooth RoCoF-proportional adaptation on 800 random scenarios covering step, ramp, forced-oscillation and double events. It produced no frequency-limit violation for step, ramp and oscillatory events, and 2.5 % for double events against 6.5–7.5 % for the benchmarks. It saved 7–19 % of battery energy in single events and varied its gains 50–75 times less than bang-bang control. Its main weakness is sensitivity to latency: violations rose to 2 % at 50 ms and 19.5 % at 100 ms.