What happens if grid frequency falls too far?

Generators and industrial equipment can be damaged, so protection systems disconnect them. To stop that cascading, operators automatically shed blocks of demand when frequency falls past a set threshold, which is a controlled way of preventing a wider blackout.

Do batteries really respond faster than power stations?

Yes. A battery inverter can change its output in well under a second, while a steam turbine takes tens of seconds to change its steam flow. That speed is why batteries won frequency response contracts before they were cheap enough for bulk energy storage.

Are ancillary services expensive?

They are usually a small share of total wholesale electricity costs, far less than the cost of the energy itself. They matter out of proportion to their cost because without them the rest of the system cannot run.

Ancillary Services and Frequency Response

Ancillary services are the background jobs that keep a power grid stable: holding the alternating current at its target frequency, keeping voltage within range, and keeping spare capacity ready in case a large generator trips offline. Grid operators buy these services from power stations, batteries and large electricity users, and pay for them separately from the electricity itself. Frequency response is the best known of them, because frequency is a continuous, grid wide signal of whether supply and demand match.

Frequency is a measure of balance

Most of the world runs its grids at 50 hertz, meaning the alternating current changes direction 50 times a second. North America, most of South America, Saudi Arabia, South Korea and western Japan run at 60 hertz. (Japan is unusual in having both: the east of the country is 50 Hz and the west is 60 Hz, a split inherited from equipment bought from different countries in the 1890s.)

The exact number matters less than the fact that it must stay almost constant. Across a synchronous area, every connected generator spins in step. If demand suddenly exceeds generation, that extra demand is met in the first instant by slowing the spinning machines down, and frequency falls. If generation exceeds demand, the machines speed up and frequency rises. Nobody measures the balance directly. They measure the frequency, and the frequency tells them the balance.

Grids hold frequency inside a narrow band, typically a fraction of a hertz either side of target during normal operation. Persistent drift outside that band damages equipment and, past a further threshold, triggers automatic disconnection of generators or of blocks of demand.

Inertia buys time

The spinning mass of a large turbine and generator stores kinetic energy. When a plant trips, that stored energy is released automatically, without anyone deciding anything, and it slows the fall in frequency. This is inertia, and it is the reason a grid built on big thermal and hydro plants can survive the loss of a gigawatt of supply without blacking out.

Inertia does not fix the imbalance. It buys seconds, which is time for control systems to act. Operators care about the rate of change of frequency: the faster it falls, the less time the reserves have to arrive.

Wind and solar connect through power electronics rather than through a machine spinning in lockstep with the grid, so they do not contribute inertia in the traditional way. As their share grows, grids need faster response to compensate. Britain, Ireland and South Australia hit this constraint early and now run markets specifically for very fast frequency services. Modern “grid forming” inverters can imitate the behaviour of a spinning machine closely enough to provide a synthetic equivalent, and hydro or thermal plants can be fitted as synchronous condensers, spinning without burning fuel purely to supply inertia and short circuit strength.

The layers of reserve

Frequency response works in layers, each slower and cheaper than the one before.

Containment, seconds. Automatic and local. Governors on generators and control systems on batteries sense the frequency deviation and change output within a second or two. The goal is only to stop the fall, not to correct it.
Restoration, tens of seconds to minutes. The operator’s central control system sends signals to generators to bring frequency back to target and return the containment reserve to standby. In North America this is largely automatic generation control, driven by each balancing authority’s area control error.
Replacement, minutes to an hour. Slower plant is started or rescheduled to take over, freeing the faster reserves for the next event.

Alongside these sit reserves held explicitly for contingencies: spinning reserve from plant already running and synchronised, and non spinning reserve from units that can start within roughly ten minutes. Operators size these against the largest single credible loss, usually the biggest generator or interconnector on the system.

Voltage support is the other large ancillary service. Voltage, unlike frequency, is local, so it has to be managed substation by substation using generators, capacitors and other equipment. A third service, black start, is the contracted ability to restart the grid from nothing after a total shutdown.

How batteries changed the market

Batteries are well matched to frequency work. They respond in milliseconds, they can absorb as well as inject power, and a frequency service asks only for short bursts of energy rather than hours of output. That combination let grid batteries earn revenue from frequency markets years before they were cheap enough for bulk storage, and several early projects, including the first large battery in South Australia, paid for themselves mainly this way.

The result in several markets has been a collapse in frequency service prices, because a handful of batteries can supply what previously required many partly loaded thermal plants running inefficiently just to keep headroom. In Texas, where wind output swings quickly and the grid has limited connections to its neighbours, fast responding batteries have become a routine part of how the system is balanced.

There is a limit. A battery holding reserve is not selling energy, and a battery that has just discharged for frequency support needs to recharge. Operators therefore watch the state of charge of the fleet, and increasingly require providers to prove they can sustain a response for a stated number of minutes.

Demand response does the same job from the other side

Balance can be restored by reducing demand as easily as by raising supply, and demand is often faster. Aluminium smelters, electrolysers, water treatment plants, cold stores and data centres can all cut load briefly without harm. So can large numbers of small devices: electric water heaters, heat pumps, freezers and electric vehicle chargers can be built to sense frequency themselves and back off within a second, with no communication network involved.

Aggregators bundle thousands of such devices and sell the combined response to the operator. For a household the effect is invisible, a few minutes of a water heater pausing. For the grid it is the same as a power station appearing on demand.

Why any of this shows up in the data

Ancillary services rarely appear in generation statistics, because the energy involved is tiny. They appear instead in system costs, in market rules, and in the technical limits an operator places on how much non synchronous generation it will accept at once. When a country debates how far it can push wind and solar, the binding constraint is often inertia and reserve rather than annual energy, which is why a grid can sit comfortably at a high annual renewable share and still refuse to run a particular hour without some conventional plant online.