Episode 1 — Grid-Forming Inverters & Virtual Synchronous Machines

"Who Holds the Frequency When Nothing Is Spinning?"

Pillar: Power & Energy Systems · Runtime ~14 min · ~2,050 words · Elena & Marcus

MARCUS: Welcome to the course. This isn't a news show and it isn't a textbook read out loud — it's a lesson. By the end of each episode you should understand a piece of modern electrical engineering well enough to explain it to someone else. I'm Marcus, and I mostly ask the questions you're already thinking. Doing the actual teaching is Elena.

ELENA: Glad to be here. Episode one, and I want to start with a question that sounds simple and isn't: when you flip on a load somewhere in the country, what physically keeps the grid at sixty hertz in that first half-second? Hold that question. By the end you'll have a graduate-level answer, and you'll understand why that answer is being rewritten right now.

MARCUS: So today's objective, concretely?

ELENA: Three things. One — re-derive why frequency has always been tied to spinning mass. Two — understand what "low inertia" really means as renewables take over. Three — understand the actual control distinction that the whole field now turns on: grid-following versus grid-forming inverters, and the virtual synchronous machine.

The Refresher: Where Frequency Comes From

ELENA: Let's rebuild the foundation. The classical grid is dominated by synchronous generators — big rotating machines in coal, gas, nuclear, hydro plants. The electrical frequency at the terminals is rigidly tied to the mechanical speed of the rotor through the pole-pair count. Sixty hertz electrical is the rotor physically spinning at a fixed mechanical rate. So grid frequency is, very literally, a tachometer reading on a few thousand enormous synchronized flywheels.

MARCUS: And they're all locked to each other.

ELENA: Synchronously, yes — that's the word doing the work. Now the governing equation, the one I want in your hands today, is the swing equation. In its compact form:

M · dω/dt = P_mechanical − P_electrical

where ω is rotor speed, P_mech is the turbine's input power, P_elec is the electrical power being pulled out, and M is proportional to the machine's inertia. We usually normalize inertia as H — the inertia constant — the number of seconds the stored kinetic energy could supply rated power. For big thermal units, H is on the order of two to nine seconds.

MARCUS: So walk me through what the swing equation says physically.

ELENA: Suppose a big load switches on, or a generator trips. Suddenly P_electrical exceeds P_mechanical — the grid is demanding more than the turbines are supplying. The equation says dω/dt goes negative: the rotors slow down. Where does the extra energy come from in that instant? Out of the kinetic energy of the spinning mass. The flywheels donate inertia, automatically, with no controller, no communication, at the speed of physics. Frequency sags, and the rate it sags — RoCoF, rate of change of frequency — is set by how much total inertia is on the system.

MARCUS: And then the governors catch it.

ELENA: Right, that's the second layer. Governors sense the speed droop and open up the turbines — that's droop control, a deliberate proportional relationship between frequency and power output so multiple machines share the load gracefully. But notice the ordering. First inertia slows the fall, buying you seconds. Then governors arrest it and restore. Inertia is the flywheel that resists the change; droop is the spring that pushes back. And one more relation to keep handy — how real power actually flows between two points on the grid:

P = (V₁ · V₂ / X) · sin δ

Power transfer is governed by the angle δ between two voltage phasors across a reactance X. Generators speed up or slow down to set that angle. Frequency, angle, and power are one coupled system. Lock that in, because inverters are about to play exactly the same game by completely different means.

What Changed: The Inertia Problem

ELENA: Here's the shift since your 2013 baseline. Back then, solar and wind were a small slice riding on a grid still dominated by spinning machines. Today, on a sunny, windy afternoon, entire regions can run with the majority of their power coming through inverters — from PV, wind, and batteries. And an inverter, by default, has no rotating mass. No flywheel. No kinetic energy to donate.

MARCUS: So the shock absorber is disappearing.

ELENA: Exactly the right way to say it. As synchronous machines get displaced, total system inertia falls. Go back to the swing equation: smaller M means a larger dω/dt for the same disturbance. RoCoF gets steeper. Frequency falls faster and dips lower before anything can react. And protection relays watch RoCoF — if it's too steep, they can trip more generation, and you get a cascade.

MARCUS: Is this theoretical or has it actually bitten?

ELENA: It's bitten. The 2016 South Australia blackout and the August 2019 event in Great Britain both have low-inertia, fast-RoCoF fingerprints in their post-mortems. This is the central anxiety of modern power systems: we are removing the system's natural inertia faster than we're replacing its function. So the engineering question becomes — can an inverter be made to behave like it has inertia, even though it doesn't? That's where the real content of this episode lives.

The Mechanism: Following vs. Forming

ELENA: There are two fundamentally different ways an inverter can connect to the grid, and the entire field organizes around this split. The first is grid-following. The second is grid-forming. Let me teach them as control archetypes.

MARCUS: Start with grid-following, since that's what's mostly out there.

ELENA: Grid-following — sometimes called grid-feeding. The inverter's job is to inject a commanded amount of current at the grid's voltage angle. To do that it has to know the angle, so at its heart is a phase-locked loop. The PLL measures the grid voltage, locks onto its phase, and the controller pushes current in step with it. Behaviorally, a grid-following inverter is a controlled current source. It assumes there's a stable voltage waveform out there to lock onto and follow.

MARCUS: Which is fine as long as something else is making that waveform.

ELENA: That's the whole vulnerability in one sentence. A grid-follower is a guest. It needs a strong, stiff grid — formally, a high short-circuit ratio — to lean on. Push the penetration up so that most sources are grid-following current sources, and you get a philosophical problem: everyone is following, and no one is leading. The voltage angle they all lock onto becomes weakly defined. The PLLs start fighting the very grid they're measuring, and in a weak grid that feedback can go unstable. You've all heard the intuition: a room full of people all trying to match each other's pace, with no metronome.

MARCUS: So grid-forming is the metronome.

ELENA: Grid-forming is the metronome. A grid-forming inverter does not lock to the grid's angle — it imposes its own. Behaviorally it's a controlled voltage source behind an impedance. It sets its own internal voltage magnitude and phase angle, and then real and reactive power flow naturally according to the P = (V₁V₂/X)·sin δ relationship we established — exactly like a synchronous generator setting its rotor angle. Because it defines the voltage itself, a grid-former can energize a dead network — black-start — and it can run an island with no other source present. It leads instead of follows.

MARCUS: And that's what lets it provide inertia?

ELENA: That's what makes inertia possible — but you still have to program the behavior. Which brings us to the centerpiece: the virtual synchronous machine.

The Virtual Synchronous Machine

ELENA: Here's the elegant idea. We spent Part 1 deriving the swing equation that governs a real generator's rotor. A grid-forming inverter has a fast processor and full control over its output voltage. So: take the swing equation — M·dω/dt = P_in − P_out, plus a damping term — and implement it in firmware as the control law that decides the inverter's output angle. The inverter has no physical rotor, but its controller solves the rotor's equation of motion in real time and acts it out. We call it a virtual synchronous machine, or a synchronverter.

MARCUS: So the math that used to be physics is now software.

ELENA: That is exactly the conceptual leap, and it's worth sitting with. Inertia used to be a property of steel and angular momentum — you got it for free, and you couldn't change it. Now it's a control parameter. You can dial the virtual inertia constant H up or down in a config file. You can even make it larger than any real machine, or schedule it by time of day. Inertia became firmware.

MARCUS: But there's no free lunch. If the software demands inertial energy, something physical has to deliver it.

ELENA: Crucial point — you've found the catch. When a real generator donates inertia, the energy comes from its kinetic store. When a virtual synchronous machine "donates inertia," the energy has to come from the DC side — from a battery, or headroom on a solar array, or a capacitor. The control law writes the check; the energy storage cashes it. That's precisely why grid-forming and battery energy storage are joined at the hip: the battery is the kinetic energy the inverter doesn't physically have. The controller emulates the flywheel; the battery is the flywheel's energy.

MARCUS: Are virtual synchronous machines the only way to do grid-forming?

ELENA: No, and I'll name the family so you can read the literature. Droop control is the simplest grid-forming law — a direct frequency-power and voltage-reactive relationship. Virtual synchronous machine control adds the inertial dynamics by emulating the full swing equation. And there's a newer approach, virtual oscillator control — dispatchable virtual oscillators, dVOC — based on nonlinear oscillator theory that synchronizes a population of inverters with provable stability and very fast response. Droop, VSM, and virtual oscillators — that's the modern grid-forming toolkit. They differ in how they generate the angle, but all three make the inverter a voltage source that leads.

Industry Standard Today

MARCUS: So where does this actually stand in real deployments and standards?

ELENA: It's moved from research to mandate, fast. On the standards side, the anchor in North America is IEEE 2800, published in 2022 — interconnection requirements for inverter-based resources, the rulebook for how big wind, solar, and storage plants must behave on the grid, including ride-through and frequency response. There's an active push, through groups like the UNIFI Consortium, to specify grid-forming behavior concretely. Internationally, grid operators in regions with high renewable penetration — Australia, Great Britain, Ireland, parts of Europe — are now requiring grid-forming capability on new large-scale batteries, not just allowing it.

MARCUS: Any flagship example to anchor it?

ELENA: South Australia is the standard reference. The Hornsdale Power Reserve went in in 2017, right after the 2016 blackout, as a grid-following battery selling fast frequency response. Around 2022 it was upgraded to run in grid-forming mode, a virtual-machine mode providing synthetic inertia, and newer South Australian batteries are being built grid-forming from day one. The place that had the 2016 scare became the template for the fix. That's the arc to remember: the grid lost its flywheels, and we taught power electronics to be flywheels.

Self-Test

MARCUS: Let's turn this into skill. Give the listener something to actually work.

ELENA: Two exercises. First, the back-of-envelope one. Take a power system with total inertia such that the combined H is, say, four seconds at some rated capacity, and imagine a sudden generation-loss event of a known per-unit size. Use the swing equation to estimate the initial RoCoF — dω/dt right at the instant of the disturbance, before any governor or inverter responds. Then cut the system inertia in half, modeling renewables displacing spinning machines, and recompute. You'll see RoCoF double. That doubling, in one calculation, is the entire inertia crisis.

MARCUS: And the second?

ELENA: Conceptual but it's the one that proves you understood the episode. Explain, in your own words, why a grid-following inverter with a phase-locked loop becomes unstable as the short-circuit ratio drops — as the grid gets weak — while a grid-forming inverter does not. If you can articulate that the follower depends on a stiff external voltage to lock onto, and the grid-former creates the voltage reference itself, you've got the core idea cold.

MARCUS: Next episode we cross into signal processing and communications — OFDM, the trick that makes a brutal multipath channel simple, and how 5G made that waveform shape-shift with numerology. Same lesson structure, new pillar.

ELENA: And notice the move we made today: when the physics stopped providing inertia, we changed where it lives — from steel into software. You'll see that kind of move again next time, when a change of basis makes a brutal channel diagonal.