Mastering grid forming vs grid following in real-time testing

Key Takeaways

• Compliance and stability evidence for a grid forming inverter should come from one model, one controller and one set of sequences.
• IEEE 1547 sequences confirm terminal performance and leave phase jump response, current limiting and weak grid margin to additional tests.
• An EMT model with the production controller in the loop, fed by wide-area boundary conditions, lets the full sequence set be rerun on every firmware release.

Grid forming inverter testing does its job when one test campaign proves that the inverter stays stable on a weak grid and passes every sequence its interconnection standard requires.

Those two goals are usually handled by separate teams with separate models, and that’s where problems start. A grid forming inverter behaves as a voltage source behind an impedance, so its response to a fault or a phase jump comes from control code that a nameplate can’t describe.

The cost of skipping that step is already on record. Across 10 large disturbances since 2016, NERC counted nearly 15,000 MW of unexpected output reduction from inverter-based resources, and the same review found model and protection-setting gaps across the plants it surveyed. Compliance evidence and stability evidence should come from one model, one controller and one set of sequences.

What grid forming inverter testing has to prove

Grid forming inverter testing has to prove three things. The inverter holds its voltage and frequency reference across its full operating range, it stays stable when current limiting takes over during a fault, and it meets each performance sequence in the applicable interconnection standard. A test plan that covers only one of the three leaves a gap.

Picture a battery inverter running virtual synchronous machine control at half its rated power when the grid voltage angle jumps by 30 degrees. A voltage source responds within the first cycle, pushing current to oppose the jump. Its semiconductors can only carry a small margin above rated current, so the limiter engages almost at once. What the controller does over the next few cycles decides the outcome.

That’s why pass criteria need numbers before the first run. You’ll want a defined response time, a damping target for the power oscillation that follows, and a rule for how quickly the inverter must leave current limiting. A criterion that only asks the inverter to remain stable can’t be audited.

How grid forming testing differs from grid following testing

The main difference between grid forming and grid following testing is the reference each inverter depends on. A grid following inverter measures grid voltage through a phase-locked loop and injects current against it, so its tests check tracking. A grid forming inverter sets its own voltage and frequency, so its tests check what happens when the grid moves away from that reference or disappears.

Islanding shows the split clearly. IEEE 1547 expects a unit to detect an unintentional island and cease to energize within 2 seconds. A grid forming plant approved to carry an intentional island must instead hold voltage for the loads left behind. One event produces two opposite pass criteria, and your test plan has to state which one applies at each site.

Where IEEE 1547 testing stops for grid forming inverters

IEEE 1547 testing for grid forming inverters confirms the same terminal performance required of any distribution-connected resource. IEEE 1547-2018 sets the requirements, and IEEE 1547.1-2020 defines the test procedures for ride-through, voltage regulation, frequency response, islanding and interoperability. Neither document scores the qualities that make an inverter grid forming.

Take a voltage ride-through sequence. The unit stays connected through the sag, the trip timers are respected and the result is a pass. Nothing in that record tells you if the inverter kept acting as a voltage source or dropped into a current-controlled mode with different stability properties. 

“Two firmware builds with very different weak grid margins can produce the same certificate.”

Transmission-connected plants answer to IEEE 2800-2022 and to the grid code of the system operator, and several operators now publish grid forming specifications alongside those. These add the missing checks, including phase jump response, performance at low short circuit ratio, and operation after the last synchronous machine disconnects.

Treat IEEE 1547 as the floor. It’s necessary evidence, and it was never meant to be the whole case for stability.

Which weak grid stability testing methods expose control problems

Weak grid stability testing methods work by lowering the short circuit ratio at the connection point in steps and repeating the same disturbances at every step. The approach exposes control problems because margin shrinks gradually, so you see the oscillation grow before the inverter trips.

The stakes scale with the build-out. Battery storage, the usual host for grid forming control, accounts for 28% of the 86 GW of generating capacity that United States developers plan to add in 2026, and each of those projects is studied at the grid strength of its own connection point. A practical campaign runs 5 methods.

• Short circuit ratio sweep from a stiff grid down to the weakest credible case.
• Phase jump steps of rising size at each grid strength.
• Balanced and unbalanced faults with recovery into the weakened grid.
• Impedance scans across frequency to locate poorly damped resonances.
• Planned transitions between grid-connected and islanded operation.

Run the sweep in both directions. Droop-based grid forming control tends to be comfortable at a low short circuit ratio and can lose damping on a very stiff grid, which is the reverse of what grid following experience teaches. Teams that stop at the weak end don’t find that until commissioning.

Why compliance sequences need an electromagnetic transient model

Compliance sequences need an electromagnetic transient (EMT) model because the events that decide a grid forming result last a few milliseconds. Current limiting, phase jump response and fault recovery all happen inside one or two cycles. 

“A phasor model averages across that window, so it reports a clean result for behaviour it never calculated.”

Controller hardware in the loop puts this to work. The production controller connects to a simulated feeder, transformer and converter stage that solve in real time, and a script steps through each ride-through and disturbance sequence. The firmware under test is the firmware that ships.

The payoff is repetition. A full sequence set can be rerun overnight after every firmware change, at every grid strength, with identical initial conditions.

There’s a limit you should plan around. Certification type tests still happen on physical equipment, and the simulated power stage has to be validated against measurements before anyone trusts its margins. The model earns its place when it narrows the laboratory campaign to the cases that matter.

How wide-area context changes a grid forming test result

Wide-area context changes a grid forming test result because a single inverter tested against an ideal source never meets its neighbours. Actual disturbances arrive as frequency swings, voltage angle shifts and slow oscillations shaped by hundreds of machines and other inverters. A controller that looks well damped against a fixed source can interact badly once those are present.

Consider a 200 MW storage plant studied for the loss of a large generator two substations away. A phasor-domain model of the region produces the frequency decline and angle swing at the plant boundary, and the detailed model replays that trajectory against the controller. With OPAL-RT tools the work splits naturally, with ePHASORSIM carrying the wide-area network and HYPERSIM carrying the EMT detail around the inverter. You get boundary conditions a grid code study will recognize, plus the sub-cycle detail that study can’t resolve.

What a defensible grid forming compliance record looks like

A defensible grid forming compliance record links every requirement to a sequence, a waveform and a model version. Anyone reviewing it can pick a clause from IEEE 1547 or the grid code, find the test that covers it, and see the controller’s response at each grid strength.

That record is built through habit more than heroics. Teams that rerun the whole set on each firmware release catch the tuning change that fixed a weak grid oscillation and quietly broke ride-through. Teams that test once before commissioning usually learn about it from a utility.

We’d argue the discipline matters more than the tooling, though the tooling has to keep up. Engineers working with OPAL-RT simulators tend to treat the model as a long-lived asset that’s validated, versioned and reused from design through commissioning. It’s the practice that lets you sign a compliance report and still sleep when the grid gets weaker than the study assumed.