Grid-Forming vs Grid-Following Inverters in BESS

At a glance

GFM vs GFL at a glance. Requirements are still evolving — always cite the specific grid-code edition and year.

Voltage source or current source: the defining difference

A grid-following inverter is a current source. It measures the grid voltage waveform with a phase-locked loop (PLL) and injects current at a commanded magnitude and phase relative to that reference. It needs a stable voltage to lock onto — remove the grid and it has nothing to follow. Almost all utility-scale solar and storage inverters deployed to date are GFL.

A grid-forming inverter is a voltage source behind a coupling impedance. It holds an internal voltage magnitude and phase (frequency) reference and lets current flow as the network demands, much as a synchronous generator does. Because it defines the waveform rather than chasing it, a GFM can establish voltage on a dead network, run an island, and share load with other sources.

The practical rule: a GFL asks 'what is the grid doing, and how much current should I inject?'; a GFM asserts 'this is what the voltage and frequency will be, and I will hold them.'

How a grid-forming inverter sets frequency and voltage: droop

A GFM does not hold one rigid frequency — it droops. Its frequency reference falls slightly as it exports more active power (f–P droop), and its voltage reference falls as it supplies more reactive power (V–Q droop). A dead-band keeps it from reacting to trivial deviations, and the slope sets how aggressively it shares load.

Droop is what lets many voltage sources run in parallel without fighting. Because each unit lowers its frequency the same way under load, they share power in proportion to their droop slopes — the identical mechanism that let synchronous machines parallel for a century. As synchronous generation retires, GFM BESS supplies the reference the grid once took from spinning machines. The droop visual shows the operating point sliding along the f–P and V–Q curves as dispatch changes.

Synthetic inertia and RoCoF

Neither inverter type has spinning mass, so neither provides true mechanical inertia. But after a sudden generation loss they behave very differently. On a low-inertia system, frequency falls fast — a steep rate of change of frequency (RoCoF) toward a deep nadir that can trigger under-frequency load shedding (UFLS).

A GFM responds to a frequency deviation inherently and within milliseconds, because its power output follows directly from the angle across its coupling impedance — there is no measurement loop in the path. This synthetic (or virtual) inertia flattens the initial RoCoF and lifts the nadir. A GFL can also deliver fast frequency response, but only after its PLL and control loop detect the deviation, adding delay. Neither replaces mechanical inertia; GFM emulates it most closely. The RoCoF visual shows how inertia and fast response together decide whether frequency recovers before the UFLS threshold.

Behaviour during a grid fault: current saturation vs trip

This is where the two architectures diverge most. A GFL inverter controls current directly, so during a voltage sag it can be commanded to reduce current or inject reactive current, then either ride through per the grid code or trip on its protection thresholds. Its current is inherently bounded because current is the quantity it regulates.

A GFM holds voltage, so in a fault it tries to source whatever current that voltage demands across the now very low fault impedance. That request can far exceed the converter rating, so the PCS hard-limits — saturates — the current at roughly 1.1–1.2 pu of rated. Counterintuitively, a strong, low-impedance grid saturates a GFM sooner than a weak one, because I = ΔV/Z and Z is tiny. This current headroom is the single biggest hardware constraint on GFM, and the reason a converter cannot mimic the large fault infeed of a synchronous machine. The saturation visual animates the strong-grid current request clipping before the voltage reference is reached.

Black-start capability

Only a voltage source can energise a dead network, so black start is a GFM-only capability. Starting from zero volts, a GFM raises voltage and frequency on a soft V/f ramp, energises transformers and cables, and picks up load blocks one at a time.

The constraint, again, is current. Close a breaker into a transformer at full voltage and the core saturates past its roughly 1.15 pu flux knee, demanding 5–10x rated magnetising inrush — current a converter limited to about 1.2–2.0 pu cannot supply, so it distorts its voltage or trips. The soft ramp raises volt-seconds slowly so core flux never crosses the knee and the inrush is never demanded in the first place. Each load block must fit the headroom between served load and the current limit, or the island collapses. The black-start visual traces bus voltage, delivered versus demanded current, and core flux as blocks are picked up.

When is each required? IEEE 2800, GC0137, AEMO

For most interconnections today, GFL remains the default and is sufficient on a strong grid. IEEE 2800-2022, the US standard for interconnecting inverter-based resources to transmission systems, sets ride-through and frequency/voltage support requirements that GFL can meet; it specifies capabilities rather than mandating grid-forming outright.

Grid-forming is increasingly required where inverters dominate and system strength is low. In Great Britain, National Grid ESO's Grid Code modification GC0137 introduced a Grid Forming specification for new connections. In Australia, AEMO has published grid-forming specifications and guidance and is driving GFM through connection conditions in weak parts of the NEM, with several large storage projects procured specifically for grid-forming capability. The direction of travel is consistent: as synchronous plant retires, more of the fleet must form the grid rather than follow it. Always cite the specific edition and year — requirements are still evolving.