Two inverters, the same hardware, a different question

Take two inverters, identical in every physical respect: same bridge, same filter, same current sensors, and connect each one to the grid through its own current controller. Give the first one a power reference and a PLL. Give the second one nothing but a voltage magnitude and a frequency to hold. Both will sit there quietly exporting power. If you only ever watch them in steady state, you cannot tell them apart.

Now let the grid voltage jump: a fault clears somewhere upstream, or a large load switches in, and the phasor at the terminals shifts. The first inverter does nothing for a moment: its current reference has not changed, so its current does not change, and the voltage at its own terminals is whatever the network makes it be. The second inverter’s voltage does not change: it was already holding its own internal voltage phasor, and now the network’s current through it changes almost instantly to reconcile the difference.

That is the entire distinction, and it is worth being precise about, because the usual one-line version (“grid-following is a current source, grid-forming is a voltage source”) is the right place to start but not quite a definition on its own.

Why the equivalent-circuit slogan is not quite enough

Every linear one-port electrical source has both a Norton (current source with a parallel impedance) and a Thevenin (voltage source with a series impedance) equivalent, and the two are mathematically interchangeable. So “current source” and “voltage source” cannot, by themselves, be the definition of two genuinely different things: any source can be redrawn either way. A review of grid-forming control makes exactly this point:

“It is worth noting that this representation might erroneously resemble the definition of a Norton or a Thevenin equivalent, which are theoretically interchangeable; yet it does emphasize the fact that GFL converters achieve their purposes of power injection or voltage regulation by controlling the injected currents, while the GFM converter regulates the power by controlling directly the voltage at its output terminals.”¹

So the real distinction is not about the equivalent circuit you happen to draw. It is about causality inside the controller: which quantity is an independent state the controller decides, and which quantity is left as an output for the network to determine.

• Grid-following. The controller’s independent variable is the measured terminal angle, \(\theta_{PLL}\), tracked by a phase-locked loop. From that measurement and a power reference it computes a current reference \(i^* = f(P^*, Q^*, \theta_{PLL})\), and the current loop makes the real current follow it. Causality runs voltage in, current out. The inverter’s own angle is not a thing it has: it is slaved to whatever angle it measures.
• Grid-forming. The controller’s independent variables are an internal angle \(\theta\) and magnitude \(E\), produced by the outer loop itself (Chapters 4 and 6 build exactly how). It imposes a terminal voltage reference behind an impedance, and the current is whatever that imposed voltage and the network happen to produce. Causality runs voltage out, current in.

A statement about a timescale, not about all time

No real grid-forming inverter is a voltage source forever. It is a voltage source inside the bandwidth of its own voltage loop, and only below its current limit. Push past the current limit, which Chapter 8 is entirely about, and the inverter is forced to behave like a current source again, whether its outer loop still thinks it is forming a voltage or not. The honest definition therefore pins a timescale: a grid-forming inverter holds its internal voltage phasor approximately fixed for the first few milliseconds after a disturbance: the sub-transient to transient window, because its angle and magnitude are internal states rather than measurements, not because it is magically immune to overload.

What actually happens in the first instant

This is the picture worth carrying through the rest of the book. Take a step change in the grid voltage phasor and watch each inverter’s very first reaction, before either controller has had time to do anything clever:

“Because of its inherent current source behavior, the instantaneous reaction of the GFL converter is to maintain the current phasor \(I_g\) constant in terms of magnitude and phase… According to its intrinsic behavior of a voltage source behind impedance, the internal voltage phasor \(E\) of the [GFM] converter is initially not affected by the perturbation, causing an almost instantaneous variation of the phasor \(I_g\).”²

The grid-following inverter contributes nothing in the first instant: it structurally cannot, because its response is gated by how fast its PLL can detect the new angle. The grid-forming inverter’s current changes almost immediately, because nothing in its controller is protecting it from doing so: its voltage is fixed, so Ohm’s law across its own impedance decides the current for it. That prompt reaction is exactly what a system operator wants during a disturbance. It is also, in the same paragraph of the same review, exactly what creates the problem this book returns to twice more:

“…depending on the magnitude of the perturbation and on the characteristics of the system, this behavior might cause a rapid growth of the converter currents, hence jeopardizing the converter hardware components.”³

Hold that sentence. Chapter 4 will meet it again the moment droop control is introduced (a droop controller commands a voltage, and nothing in a voltage command contains a current at all, so nothing in it can stop one), and Chapter 8 is built entirely around solving it properly.

The stability duality: each type fails in the other’s comfort zone

Grid-following inverters are well known to struggle as the grid gets weaker (Chapter 2 develops this in depth: the PLL’s own bandwidth interacts with the grid impedance and can go unstable). It is tempting to conclude that grid-forming inverters are simply better. They are not: they trade one failure mode for a different one:

“…GFM converters are instead suitable for weak grid applications. On the contrary, they result to be more prone to instability under stiff grid operating conditions compared to their counterpart GFL converters.”⁴

The physical reason for the grid-forming row is worth stating now, because it is one line of algebra and it will reappear as the central design constraint of Chapters 4, 6 and 9. A grid-forming inverter sets an angle, and for a source \(E\angle\delta\) feeding a grid \(V\angle 0\) through a total reactance \(X\), \[ P \approx \frac{E\,V}{X}\sin\delta. \] As the grid gets stiffer, \(X_{total} \to 0\), and the gain from angle to power grows without bound. Any loop that closes around \(P\) (which is exactly what droop and the swing equation do) sees an unbounded loop gain on a sufficiently stiff grid and rings or goes unstable. Virtual impedance (Chapter 9) exists partly to put a floor under \(X_{total}\) for precisely this reason: it is a stabiliser as much as it is a current limiter.

Both types can do the same job in steady state

One more qualification, because it stops a common confusion before it starts. In steady state, a grid-forming and a grid-following inverter can both regulate active power, reactive power, or terminal voltage: there is nothing a grid-forming inverter can do in steady state that a grid-following one cannot:

“In spite of different working principles, under steady-state operation, both GFM and GFL converters control active and reactive power injection into the grid according to the actual operating condition… Nevertheless, the main differences among the two types of converters can be identified in the reaction to a grid event.”⁵

Steady state does not distinguish them. The response to a disturbance does. Keep that framing: it is the reason every later chapter’s worked examples are about transients, faults and closing transients rather than steady-state operating points.

What this means for the rest of the book

A common shorthand for this book’s organising idea says “the voltage controller, current controller and PWM are identical across every method.” That is very close to right, but it overstates one part of it, and the correction matters enough to fix now, before it is repeated. A grid-following inverter, as described above, has no voltage controller at all: its cascade is PLL, then a power or current reference, then the current controller, then PWM. There is no block anywhere in it that regulates a voltage.

So the shared blocks across all three control methods this book covers are the current controller and the modulator, full stop. The voltage controller is shared between droop and VSG only, and its presence is precisely what makes a converter grid-forming in the first place. That is a stronger, more useful statement than the original, because it now explains two transitions with one sentence instead of one: