Demand Response: The Cheapest Power Station Is a Pause

Industry Got There First

Aluminium smelting, electric arc steelmaking, chlorine production and industrial air separation share a useful property: they consume enormous amounts of electricity and can tolerate short interruptions. A smelter drawing several hundred megawatts can reduce load within seconds without damaging the potline, provided the interruption is brief.

Utilities recognised this decades ago and wrote interruptible tariffs. The customer receives electricity at a discount and in exchange grants the operator the right to cut supply for a defined number of hours a year with defined notice. For the operator it is capacity without construction. For the smelter it is a permanent reduction in an input cost that dominates its economics.

These arrangements are unglamorous and they work. In several European systems industrial interruptible load remains the largest single block of demand-side capacity, dispatched reliably because a handful of large sites are simpler to coordinate than a million small ones.

The limitation is that the pool is finite. There are only so many smelters, and their willingness to interrupt is bounded by production schedules. Growth in demand response has to come from smaller and more numerous loads, which is where the difficulty begins.

Why Thermal Loads Are the Easy Ones

The loads that shift most readily are those that store energy as temperature. A freezer warehouse holding minus twenty-two degrees can drift to minus twenty over an hour without threatening the goods, and in doing so it has effectively discharged a battery. A domestic hot water cylinder heated at three in the morning rather than seven in the evening delivers exactly the same service to the household.

This matters because thermal loads are an enormous share of electricity demand and growing. Heat pumps, air conditioning and water heating together dominate residential consumption in most climates, and a building's own thermal mass provides storage at no additional cost. Pre-heating a well-insulated house by one degree before a peak is invisible to the occupants and removes the load entirely from the critical hour.

Electric vehicles are the largest new opportunity. A car plugged in at seven in the evening and needed at seven the next morning has twelve hours to absorb four hours of charging. Simply deferring that charge past the evening peak, with no change to when the car is ready, is one of the cheapest grid services available - and unlike most flexibility, it grows automatically as the fleet grows.

What all of these share is that the service delivered to the user is unchanged. Nobody experiences a colder house, a warmer freezer, or a car with less range. The shift is invisible, which is precisely why it is acceptable at scale.

Why the Household Version Keeps Disappointing

Utilities have run time-of-use tariffs for decades on the theory that price signals change behaviour. Results have been consistent: households do respond, by a few percent, and then the response decays over months as novelty fades.

The reason is simple enough. The sums involved are small relative to the attention required. Saving thirty cents by running a dishwasher at eleven at night is not worth reorganising an evening around, and no amount of tariff design changes that ratio.

What does work is automation. When a heat pump, water heater or car charger responds to a signal without the household deciding anything, participation rates rise several-fold and persist, because there is nothing to sustain. The behavioural problem disappears once behaviour is not required.

This is why the aggregator model emerged. A company contracts with thousands of households, installs or communicates with controllable devices, and bids the combined flexibility into markets as a single resource - a virtual power plant. The household sees a fixed annual payment or a lower tariff and makes no decisions at all. Several European markets now have aggregators dispatching hundreds of megawatts this way, which is real, though still small next to what the theoretical potential suggested.

The remaining obstacles are administrative rather than technical: metering standards, rules on who may control a device, and settlement arrangements that establish how a reduction is measured against what would otherwise have been consumed. That last problem - measuring something that did not happen - is genuinely hard and is where most disputes arise.

What It Is Worth

Demand response competes with generation on equal terms in most modern markets. It bids into capacity auctions, provides frequency response, and relieves congestion on constrained parts of the network. In each case the avoided cost is what makes it valuable.

Per avoided kilowatt at peak, it is usually cheaper than building anything. A water heater that already exists costs nothing to install, and a control signal costs nothing to send. Against a peaking plant that must be built and maintained for a few hundred hours of use a year, the comparison is not close.

Its limitation is duration and repeatability. A freezer can drift for an hour, not for a day. A smelter can interrupt a handful of times a season, not daily. Demand response handles sharp, short peaks well and prolonged shortfalls poorly, which is why it complements rather than replaces storage and firm generation.

The practical view held by most system operators is that flexibility should be taken in ascending order of cost, and demand response sits at the bottom of that stack for the first few hours. Using it first is what makes everything built above it smaller.