Mastering grid forming vs grid following in real-time testing

Key Takeaways

• Grid-forming control now starts the right technical conversation for weak grids, high inverter share, and disturbance recovery.
• Grid-following control still fits strong systems, but it works best as the baseline reference for comparison rather than the default assumption.
• Closed-loop real-time testing matters because timing, protection, and plant interaction will decide if control behaviour holds up outside offline models.

Grid-forming control should be your default starting point for inverter-based grid testing.

Utilities now ask if an inverter can hold voltage and frequency steady when the grid stops acting like a stiff source. Solar PV supplied 5.4% of global electricity in 2023. That shift means a simple definition is no longer enough. You need proof that the controller stays stable during weak-grid operation, disturbances, and recovery.

Grid-following control still fits strong systems, but it now serves as the baseline you measure against. Grid-forming control matters when black start, islanding, fault recovery, and weak-grid operation all sit in the same study queue. You need a comparison that leads straight to test design, because control claims mean little until timing, protection, and plant interactions are exposed on the bench. That is where definition gives way to validation.

A grid-forming inverter establishes the electrical reference

A grid-forming inverter sets its own voltage magnitude, phase angle, and frequency reference, then regulates power around that reference. You use it when the inverter must behave like a voltage source for the local system. That control posture lets the plant hold the bus steady while other devices react.

A campus microgrid after a feeder trip shows the point clearly. One battery inverter can set the local waveform, pick up building loads, and hold frequency while smaller solar inverters reconnect. A black start sequence uses the same behaviour. The plant needs a source that creates a stable reference before follower devices can contribute.

That role changes how you test the controller. You are no longer asking only if current tracks a command. You are asking if voltage recovers cleanly, if frequency settles after a load step, and if power sharing stays orderly when another source enters the bus. 

“Grid-forming control earns its value when the reference itself becomes part of the disturbance.”

A grid-following inverter depends on an external reference

A grid-following inverter measures an existing grid waveform and synchronizes to it before injecting current. You use it on strong systems where another source already keeps voltage and frequency stable. That makes it effective for power conversion, yet dependent on a reference it does not create.

A utility-scale solar plant tied to a strong transmission bus is a common example. The inverter locks to the grid voltage, follows dispatch commands, and manages active and reactive current within its limits. That works well when the upstream system stays firm. The controller does not need to establish the bus because the grid already does that job.

Limits appear when the reference degrades. A weak point of interconnection, a large phase jump, or a nearby fault can leave the control loop chasing a waveform that is moving too much or disappearing altogether. That is why grid-following control now reads as the legacy baseline in many utility studies. It still belongs on the grid, but it will not answer every stability question you’re being asked to solve.

Grid forming vs grid following begins with grid strength

The main difference between grid-forming and grid-following control is the source of the voltage reference. Grid-forming control establishes and stabilizes that reference during weak or islanded conditions. Grid-following control waits for a usable waveform, then injects current into it. That split defines the first test question you should ask.

A stiff transmission bus gives a follower ample support, so synchronization margin and current limiting become the main concerns. A remote feeder with high impedance changes the picture. Voltage can swing more with each power change, so the plant needs a source that steadies the bus instead of merely reacting to it. If you’re unsure where to start, look at system strength before you look at control fashion.

That distinction also shapes plant architecture. A mixed site with solar and storage often assigns one unit to form the grid while the rest follow. Control selection then becomes a system question, not a device question. The useful comparison is shown below.

High inverter share makes grid forming the default

High inverter share shifts the default question from synchronization to system support when the waveform weakens. That is why grid-forming control now anchors many new studies. Grid-following control still matters on stiff systems, yet it no longer frames the whole technical problem.

Utilities do not need a distant scenario to justify that shift. United States interconnection queues held more than 2,600 GW of solar, storage, and wind at the end of 2023. A new solar-plus-storage project entering a weak area of the grid will be judged on voltage support, disturbance recovery, and plant interaction from the start. That is a grid-forming conversation even before equipment is ordered.

This does not remove grid-following control from the toolbox. It places it in a clearer role. If your system has strong short-circuit support and another source owns the waveform, a follower remains a sensible fit. If the system leans on inverter-based resources to hold itself together, starting with grid-forming assumptions will save you time and keep the study aligned with the actual risk.

Fault ride-through reveals control stability under stress

Fault ride-through testing shows if a controller stays stable when voltage collapses, current limits engage, and protection logic starts to act. A grid-forming inverter must keep a usable reference through that disturbance, then recover without oscillation, latch-up, or control hunting.

A three-phase voltage sag near the point of interconnection is a good stress case. The controller will hit current limits, reactive support will surge, and the internal voltage reference will be pushed hard. Recovery after fault clearing matters as much as the fault itself. A controller that survives the dip but rings for several seconds after clearing still creates a plant-level problem.

Good testing separates behaviours that look similar on a trend plot. One design will recover with tight damping and clean power restoration. Another will show phase wobble, poor sharing with parallel units, or repeated protection resets. Those are not cosmetic flaws. They tell you how the plant will behave when the grid is already in trouble and every control loop is short on margin.

Closed-loop real-time testing reveals control limits early

Closed-loop real-time testing exposes interactions that offline studies will miss because controller code, I/O timing, and plant dynamics all run against a live power-system model. You need that setup when small delays can shift a stable design into poor damping, false trips, or failed resynchronization.

A practical bench setup connects the actual controller hardware to a digital model of the feeder, source impedance, breakers, and measurement chain. OPAL-RT is relevant here because engineers use it to execute those closed-loop cases at test speed instead of inferring behaviour from separated studies. A weak-grid event can then be repeated with the same fault timing, the same plant controls, and the same I/O path.

“Repeatability matters when you are tuning a controller that sits close to a stability limit.”

The payoff is not a prettier waveform. You find control interactions before field commissioning turns them into schedule pain. A plant-level voltage loop that looked fine in offline work can show poor damping once real sampling, signal scaling, and breaker logic are added. That early exposure is why real-time validation belongs near the start of control development, not only at the end.

A useful test plan starts with weak-grid scenarios

A useful grid-forming test plan starts with weak-grid cases because that is where control claims are easiest to disprove. You should stress voltage reference quality before moving to ideal conditions. Strong-grid tests still matter, but they won’t reveal the same control limits.

The sequence you choose keeps the bench focused and keeps your results comparable across controller revisions. Starting with weak-grid cases first keeps the baseline difficult and consistent. Repeating that order after each software update makes differences easier to spot. That structure turns the list below into a usable test routine instead of a loose checklist.

• Model a low-strength grid connection before you model an ideal bus.
• Apply balanced and unbalanced faults with several clearing times.
• Test load steps that force active and reactive current limits.
• Check transitions between grid-forming and follower plant states.
• Repeat each case with realistic measurement and breaker delays.

A storage inverter intended for remote feeder support should not start with a clean infinite-bus case. Start with a weak source, then add the disturbances that will expose loss of damping or poor fault recovery. That order makes comparison easier across software revisions because each new result is judged against the same difficult baseline. Strong-grid cases still belong later in the plan to confirm normal operating performance and handoff behaviour.

Hardware timing errors can invalidate a passing result

Hardware timing errors can make a weak controller look stable or make a stable controller look unstable. Sample delay, transport lag, and unsynchronized measurements alter the phase seen by the control loop. If you ignore them, a passing plot won’t mean the plant will behave the same on site.

A common case appears when voltage measurements arrive one cycle late while breaker status updates arrive on time. The controller then reacts to an old waveform with current commands meant for a newer system state. Another case shows up when plant-level and unit-level controls use different time bases. Recovery can look acceptable in simulation logs, yet the hardware loop will show extra overshoot or false protection action once those offsets are present.

This is why the useful comparison never stops at a definition of grid-forming and grid-following control. The judgment comes from repeatable closed-loop evidence that shows who sets the reference, who depends on it, and how the plant recovers when the grid turns difficult. OPAL-RT fits that need because it gives engineers a way to validate timing, control interaction, and disturbance response in real time, where assumptions are tested instead of preserved.