Grid inertia loss

Why does a power grid full of modern inverters lurch at a fault that older spinning plant would have absorbed unnoticed?

A large generator trips off. On a grid served by conventional plant, the frequency dips slightly and recovers; most people never know it happened. On a grid largely served by inverter-connected generation, the same loss produces a sharp excursion, and may trip protection that disconnects customers.

The inverters are not faulty, and the grid is not short of generation — the remaining plant can supply the load perfectly well. Something has been lost that is not capacity, and the loss only shows up in the first moments after a disturbance.

Start with what actually happens in the instant a generator disappears. Load is unchanged; supply has just fallen. Power cannot be stored in the network itself, so in that instant the books do not balance, and something has to make up the difference before any control system has time to respond.

On a conventional grid, something does, and it is not a decision. Every synchronous generator on the system is a large mass spinning in lockstep with the grid frequency. Those masses hold kinetic energy, and they are electromechanically coupled to the network — so when demand exceeds supply, the network draws the shortfall out of them directly. They give up energy, and because energy is coming out, they slow down. Frequency falls.

That is the key realisation: frequency falling is not the problem, it is the mechanism. The dip is the visible sign of stored energy being spent to cover the gap. And the rate at which frequency falls — the rate of change of frequency — is set by how much stored energy is available relative to the size of the loss. More spinning mass means the same shortfall produces a gentler slope.

That slope is what buys time. Governors take seconds to open valves; automated fast response takes hundreds of milliseconds at best; operators take minutes. None of them can act in the first instants. Inertia is what carries the system through the interval before any controller can respond, and it does so passively, with no measurement, no communication and no decision.

Now the inverter. A solar farm or battery connected through power electronics has no rotating mass coupled to the grid, and even a wind turbine's rotor is decoupled by its converter. Such a plant supplies exactly what its control system tells it to supply. It has no automatic, physics-driven response to a frequency change, because nothing about it is mechanically tied to grid frequency in the first place.

So displacing synchronous plant with inverter-based plant does not reduce energy or capacity. It reduces the buffer — and with a smaller buffer, the same loss produces a steeper decline and a deeper trough before the controls catch it. That is the lurch.

Two consequences follow, and both are visible in operating practice. Protection settings that were tuned for gentle slopes may now see rates of change that look, to them, like a much more serious event — so the response to a disturbance can itself become a disturbance. And the size of the largest single loss a grid can tolerate is no longer set by spare capacity but by how steep a slope the system can ride out.

This is a different question from balancing supply and demand, and the boundary is worth drawing sharply. The companion account of grid balancing is about the continuous, controlled matching of generation to load — droop response, dispatch, reserves, and the fast frequency response that inverters can be made to provide. All of that is control: something measures a deviation and acts on it. Inertia is not control. It is the passive energy store that determines what happens in the interval before any controller, however fast, has responded. The two are complementary, and conflating them leads to the claim that fast response simply replaces inertia — which is not quite right, because response cannot begin until a deviation has been detected, and inertia is what governs how bad things get inside that detection window.

Inverters can synthesise some of this, and the distinction between the two kinds matters. A grid-following inverter measures frequency and adjusts output; it can deliver very fast response, but it depends on a stable frequency reference to follow, which becomes harder to obtain as synchronous plant retires. A grid-forming inverter behaves as a voltage source with its own internal reference and can be programmed to emulate inertial response. The emulation is genuinely useful and is being deployed. Whether it is fully equivalent to physical inertia — particularly at very high inverter penetration, and during faults rather than mere generation losses — is an open engineering question, and I would treat confident claims in either direction with caution.

Batteries are not automatically the answer, though they are often presented as one. A battery has energy and can deliver it quickly, but whether it contributes anything inertial depends entirely on its inverter's control mode, not on the cells. A conventional grid-following battery contributes fast response and no inertia.

Numbers here are jurisdictional and situational. Acceptable rate-of-change-of-frequency limits, inertia floors, and the largest tolerable single loss are set by system operators, differ between grids, and change as rules are revised. Small islanded systems face this far more acutely than large interconnected ones, simply because a single loss is a larger fraction of the whole. I have quoted none of these; the applicable figures belong to the specific system.

The falsification test. If the mechanism is a passive stored-energy buffer rather than a control shortfall, then the initial rate of change of frequency after a loss should track the online synchronous inertia at that moment — and should be unaffected by how much fast response is armed, since that response has not yet acted. If the initial slope varied with the amount of fast frequency response available, inertia would not be doing what this account claims, and the problem would be one of control speed alone.