Grid balancing
A Socratic walk-through of grid balancing — reasoned out one step at a time, not lectured.
The question we started with
THE QUESTION #Why must an electricity grid match supply to demand from one second to the next?
Almost every other network we rely on has slack in it. A water main has pressure in the pipes, a warehouse has stock on the shelf. Each absorbs a mismatch between what arrives and what is taken, and nobody notices. The electricity grid has essentially none of that: every kettle switched on anywhere must be answered, within the same second, by something somewhere turning slightly harder. Why build a system with no buffer at all?
Reasoning it through
REASONING #Ask first where a buffer would live. The wires store a negligible amount of energy compared with what the system moves each second. Batteries and pumped hydro exist, but they are devices attached to the grid, not a property of it — so absent those, the network is a rigid coupling between everything generating and everything consuming.
But "has to" is doing suspicious work. What actually happens if demand exceeds generation? The energy does not simply fail to appear. It gets taken from the only other store available: the kinetic energy of everything spinning. Every large synchronous generator is a heavy rotor turning in lockstep with the grid's alternating current. Draw more than is being supplied and those rotors are dragged down — they slow, all together.
Now notice what that gives us. Rotor speed is grid frequency. So frequency is not an arbitrary specification to be maintained; it is the readout of the balance. Above nominal — 50 Hz or 60 Hz depending on where you are — means more is being generated than consumed. Below means the reverse. An operator watching frequency is watching the balance itself, continuously and everywhere at once, without metering a single customer.
How much time does the spinning mass buy? That depends on how much of it there is: the rate at which frequency falls after a sudden loss is proportional to the size of the loss and inversely proportional to the system's total inertia. Heavy synchronised mass makes frequency fall slowly, a light system makes it fall fast. Either way, seconds — not minutes.
Which is why the response is layered by speed. Inertia acts instantly and involuntarily, arresting the fall. Governors sense the speed drop and open valves within seconds, according to a droop setting, halting the decline and holding frequency somewhere below nominal. Automatic generation control then adjusts set-points over minutes to return to nominal and refill the reserves used. Underneath everything sits the last resort: automatic under-frequency load shedding, which disconnects blocks of customers past defined thresholds — in Great Britain the scheme begins at 48.8 Hz. That is not a failure; it is the system deliberately shrinking demand to match what is left. Britain's August 2019 event ran that whole ladder in under a minute and ended with about a million customers disconnected.
Where do renewables fit? Wind and solar reach the grid through inverters — power electronics, with no heavy rotor mechanically locked to system frequency. A conventional inverter contributes no inertia. So a system displacing synchronous plant with inverters has less spinning mass, frequency falls faster after a fault, and the seconds available shrink.
Notice what that argument is and is not. It is not "renewables cannot supply enough energy" — over a year the totals can be made to work. It is a claim about rates: how fast the system can be pushed off balance, and how fast back. Hence a vocabulary of ramp rates, reserve and inertia rather than terawatt-hours.
The analogy
THE ANALOGY #Think of a heavy potter's wheel driven by a foot treadle. There is no store of finished pots — what you shape is what you shape now. The wheel's mass does not do the work, but it keeps the speed steady between kicks and gives your foot time to react when your hand presses harder into the clay. Swap the stone wheel for a light one and the same press stalls it before your foot can answer.
A treadle wheel has one source and one load with a single person coordinating both, whereas a grid has thousands of each, no central mechanical connection between them, and control loops that must act on a shared electrical signal none of them owns.
Clarifying the model
THE MODEL #Frequency is a system-wide variable within a synchronous area, not a local one, so a deficit anywhere shows up everywhere. Voltage is the local variable, and confusing the two obscures why balancing is a network-scale problem while voltage control is a regional one.
Inertia, meanwhile, is not a service anyone provides on purpose; it is a by-product of using spinning machines. That is why losing it caught operators out — nobody was buying it, so nobody noticed it draining away until frequency events began evolving faster.
And the fix is not inevitably more spinning mass. Inverters can help, in two ways worth keeping separate: fast frequency response, where a battery or wind farm measures frequency and injects power within a fraction of a second, and grid-forming control, where the inverter behaves as a voltage source with a synthesised inertial characteristic rather than following the grid's existing waveform. A fast electronic response can act sooner than a turbine governor ever could, but its limits differ — it needs energy actually available to deliver, and it responds to a disturbance only once measured, whereas true inertia opposes the change at the instant it occurs. This is an active engineering and standardisation problem, not a settled one.
One misconception worth correcting: the grid is not "kept full" and drawn down like a reservoir. There is nothing in it to draw down. The nearest thing to a reserve is the momentum of the machines, and spending it shows immediately as frequency falling.
A picture of it
THE PICTURE #How to readStart at Balanced and follow the arrow taken when generation is lost. Every transition is labelled with the mechanism causing it and, where it matters, its timescale — so the diagram is a race between the falling frequency and the responses trying to catch it. The branch out of Arrested is the whole renewables question: with enough inertia and fast response the system reaches Stabilised and then Restored, while a low-inertia system can hit the shedding threshold before help arrives. Shedding is a recovery path, not a failure state; only the branch where it is too slow or too small ends in cascade.
What became clearer
WHAT CLEARED #The grid has no buffer, so it uses momentum as one: rotating mass absorbs a mismatch for a few seconds and pays for it by slowing down. That makes frequency a direct, universal readout of the supply-demand balance, and every control action a race against a clock set by how much mass is spinning. Seen this way, the challenge of variable renewables is not insufficient energy but faster demands on the system's reflexes with fewer seconds to meet them — a problem of ramp rates, reserve and inertia, answerable with fast electronics rather than only with turbines.
Where to go next
ONWARD #- How droop control lets thousands of independent generators share a load increase without communicating.
- What a grid-forming inverter actually synthesises, and why that differs from merely following the grid.
Key terms
TERMS #| Term | What it means |
|---|---|
| System inertia | the total rotational kinetic energy of synchronised machines, which resists changes in grid frequency. |
| RoCoF | rate of change of frequency, the speed at which frequency moves after an imbalance; higher when inertia is low. |
| Droop control | a governor setting that makes a generator's output rise in proportion to a fall in frequency, sharing response automatically. |
| Under-frequency load shedding | automatic disconnection of blocks of demand at defined frequency thresholds, as a last-resort rebalancing. |
| Grid-forming inverter | power electronics that establish a voltage waveform and can emulate inertial behaviour, rather than following an existing grid. |
Every term the collection defines is gathered in the glossary.