Earth Notes: On Dynamic Demand Control Value (2009)

Electricity grids already have to cope with large swings in demand, minute-by-minute, cycle-by-cycle. This is at least in part unpredictable. Alongside this there is planned and sudden unplanned disconnection or failure ('tripping') of generators. With an increasing proportion of intermittent renewable generation such as wind in the mix (overtaking hydro for the first time in 2007 in the UK for example), the task gets a little harder.

Grid Balancing

Grid operators such as the National Grid in the UK deal with this 'balancing' in a number of ways. Generators on can be on hot (eg 'frequency-response') or warm or cold standby and can respond to power shortfalls with varying degrees of speed and efficiency on the 'supply' side of the equation.

On the 'demand' side, ie how much power is required moment-by-moment, large users have time-of-day (ToD) contracts, often in half-hourly (HH) slots. They are charged more when demand (and thus wholesale price) is higher, which encourages them to move their loads to other times of day.

UK domestic users with electric water and space heating can sign up for the "Economy 7" tariff of 7 cheap night-time hours (often 1am to 8am) which is a relatively crude from of ToD pricing, It was mainly designed to ensure that the output of nuclear power stations and other baseload generation that cannot be turned down at night when demand is relatively low is actually used productively rather than being run to earth and wasted. This is not intended to respond dynamically to short-term grid stress but it does help shift load away from peak times.

(Note that domestic electricity demand is about 30% of UK total in 2008.)

Grid operators also have contracts with large commercial/industrial power users to allow instant remote/automatic disconnection of the users' load by the operator at times of grid stress/emergency, such as when a generator 'trips' and goes off-line. The user may get a regular payment proportional to the size in kW of the load and how long per day it can be disconnected for. Large cold-stores don't mind power interruptions for a while once cool enough so make good interruptible (sheddable) loads, for example.

All of these are large centralised chunks of control and power.

As I've noted elsewhere, distributed microgeneration with storage might be able to help with balancing. It could be an administrative nightmare for the grid operator to manage and monitor thousands-to-millions of such 'embedded' contributions without some kind of commercial aggregator, even if the potential super-fast response from battery would be very good.

Another distributed solution in terms of non-industrial users is to have [archive] domestic appliances that are big consumers of electricity, especially those that heat or cool, but with no desperate real-time deadlines to meet, to automatically defer some of their load when the grid is under stress.

When the mains frequency (or in more extreme cases the voltage) drops below the specified nominal value (50Hz for the UK for example) it is a clear indication that the grid is under strain. This can be used in 'dumb' but energy-hungry domestic appliances such as fridge/freezers, washing machines and dishwashers, immersion water heaters, heat pumps and air-con, maybe even electric cookers/stoves, hairdriers, etc, etc, to defer or reduce their activity (drop their instantaneous load in Watts). For example by adjusting thermostats slightly and reducing the rate of resistive heating, spreading their energy consumption over a longer time and away from the crisis. It potentially only requires a fairly simple/cheap bit of extra electronics.

The response of some grid-friendly appliances can be very fast, reducing demand for (resistive) water heating for example within a tiny fraction of a second if need be, mains cycle by mains cycle. Some other appliances, eg those that have components that should not be 'short cycled' such as heat-pumps and fridge/freezer compressors, will have a more leisurely response of the order of seconds or minutes.

Another emerging class of heavy domestic power consumers is the electric car. Operators are already working with car manufacturers on "vehicle-to-grid" (V2G) solutions where the car is grid-sensitive. Eg backing off charging when the grid is loaded unless a journey is imminent, and possibly even contributing some of its energy back to a stressed grid, on the lines of the microgen support that I mentioned above. V2G response to grid stress can probably be very fast if necessary. There has been some interesting recent discussion in the light of US presidential candidate Obama's energy policy.

Smarter Internet-connected (or radio-equipped) devices such as computers can potentially be given more sophisticated signals to trim power consumption using their already-present power-saving mechanisms to achieve the same end; to allow automatic distributed dynamic demand control of possibly millions of smaller devices, and without significantly inconveniencing the end users, and probably in fact improving reliability and reducing electricity costs. Computers can often trim some elements of consumption very quickly, eg CPU speed and load, and others will be slower such as spinning discs down.

Some operators around the world already use mechanisms like these, eg: [archive] "ripple control" of water heaters in New Zealand and South Africa, so it certainly can be done and need not be prohibitively expensive. Indeed, software installed in Internet-connected computers can be essentially free for example.

Combined with dynamic time-of-day pricing these could be powerful tools to improve grid service and admit more intermittent/renewable grid energy, allowing appliances to consume when it's available and defer when not.

In a private email 2008-08-14, David Smith, Contracts and Settlements Manager at National Grid, confirmed that NG believes there is potential economic value in harnessing dynamic demand technology:

Thus far, [potential providers that we have been talking to] are proving to be technically capable of providing a service and the providers feel they are economic at our present costs. In order to facilitate a route to market for dynamic demand we are presently looking at a number of our services and proposing a number of improvements.

Dynamic Demand replaces on-demand instant Megawatts of load-following generation in the conventional 'frequency-response' grid balancing system, often using relatively carbon-intensive part-loaded open-cycle gas turbines (OCGT), with on-demand instant 'Negawatts' of automatically-conserved power.

Monetary Value?

But what actual value might accrue from these benefits, and does it exceed the capital and other costs of the enhanced equipment, ie is it financially worth supporting this distributed grid balancing? I assume for now that distributed 'dynamic control' is not necessary to keep the lights on, and that centralised solutions can continue to take the full strain if need be (although 'dumb' electric vehicles might test that assumption rather harshly). In other words, is it cheaper or better (or at least comparably good) to use this distributed mechanism as part of the UK/EU mix rather than just doing everything centrally with 'big iron' as has been the case to now in the UK?

The National Grid publishes its balancing costs, so other than any infrastructure peak capacity savings, we know what the maximum monetary value today might be if every suitable domestic appliance magically became grid-sensitive, given an estimate of all the load/power they would sum to and the correlation of their load times with the times that balancing is needed.

These balancing costs in the UK grid are close to £500m (USD1bn) in 2008, representing possibly as much as £5 per year or 1% on the average domestic electricity bill. At a very shallow analysis the extra/capital cost of adding grid-sensitivity to domestic appliances should not add (much) more than this (£5) per year to be financially viable, allowing for the lifetime of the appliances concerned. That 'budget' doesn't seem too hard to meet implying a factory-gate price hike of maybe £1 per appliance (assuming 1:5 factory:retail price ratio, cheap ~200ppm crystal timebase, etc) if the average household replaces no more than one major appliance per year on average. Considering that the applicable appliances tend to cost in the range £100 to £600+ then an extra £5 on the purchase price that in effect might contribute to paying for itself within a year is not bad. I assume that in volume the extra electronics for basic dynamic response will cost pennies at the factory gate.

However this simplistic analysis is only really an upper bound, and does not acknowledge several important engineering considerations that will reduce the actual value of this mechanism:

• The total load of grid-sensitive appliances may not be high enough to displace all the spinning/frequency/other reserve needed for balancing, ie in terms of GW, of the order of 15-25% of UK peak or maybe very roughly 15GW.
• Not all appliances may have deferrable load all the time, ie I must charge the car and wash my clothes now to be able to make my meeting or catch my plane or whatever.
• Not all of the deferrable load may be available when balancing is needed, eg we won't all be washing our clothes when a nuclear generator trips.
• Not all of the deferrable load can be deferred instantly, eg because of short-cycling and similar constraints.
• Most of the mechanisms discussed above reduce demand when the grid is overloaded, but sometimes the reverse is needed, to absorb excess power when load is lower than expected in all or part of the network.
• The simple mechanism for 'dumb'-ish appliances based on mains frequency cannot respond differently in different parts of the UK because the mains frequency is the same throughout (synchronous), whereas transmission limitations might mean that we have excess in some parts and shortages in others.
• Whilst well-understood engineering solutions from areas such as networking (eg randomised exponential backoff and slow-start) can be used to avoid grid instability even with millions of autonomous participants, only a fraction of the response could be explicitly directed/'called'/'dispatched' by the grid operator eg in anticipation (such as home PC power consumption via the Internet), and the rest would follow its preprogrammed reactive randomised course which might make turning the supertanker that is the grid significantly more cumbersome and delicate.

We also have projections from the National Grid as to how their balancing costs will change over time, in particular to deal with more intermittent generation. Those balancing costs can reasonably be expected to rise, thus potentially increasing the value of quick-acting 'dynamic demand' by reducing the amount of inefficient 'peaking plant' and 'frequency response' plant that would have to be kept running.

Scaling up for the EU

If dynamic demand technology was to be scaled up to the entire EU, eg though support from appliance energy-efficiency labelling and VAT mechanisms, then it would be important to see if approximately the same costs/benefits apply and if overall the scheme would be valuable.

The UK experience does not directly translate to other corners of the EU; for example the Scandinavian cross-border integration of nuclear, wind and hydro potentially goes a long way to solving the balancing issue for the area as a whole even at high levels of intermittent generation. Denmark has >20% wind and does sometimes export excess (reducing hydro use) and sometimes imports heavily when winds are low; it could not easily stand on its own.

Supporting Intermittent Renewables

Distributed dynamic demand control of domestic appliances can also potentially help with the 'dull windless day' scenario that becomes more important as the fraction of wind/solar/etc generation rises by automagically letting non-critical things run slower or be deferred until more energy is available. That would mean that we absolutely would not need 100% fossil-fuel dispatchable generation to 'back up' our renewable sources even when all normal 'intermittent' generation was (freakishly) absent, since this type of mechanism could help hold down demand somewhat for hours or even days at a time if necessary, along with other extant tools at the System Operator's disposal today.

For example, come 'dull windless day', your wash might take 1 hour instead of the usual 45 minutes unless you were desperate and explicitly pressed the 'Urgent' button (possibly paying more for the electricity) to override the 'grid-friendly' feature because you had an impatient hot date arranged...

The flip-side is that the cost of the electricity for the wash could be slashed if you set the washing machine to wait a while until the grid is showing signs of excess available power by pressing the 'Economy' button.

In either case the total energy used to perform the wash may be similar, but in the 1 hour case the water is heated at half the rate/power (and maybe to a slightly lower temperature to economise still further), halving the peak load on the grid from this appliance. The extra time spent agitating the clothes while the water heats ensures that the slow economy wash is as good as a normal one. Likewise, all non-urgent washes would take 1 hour instead of 45 minutes until the wind picks up, so the lights stay on without fossil-fuel burn.

This implies that you'd not only be charged a variable amount per kWh by ToD, but also a premium for using energy (too) quickly when the grid is stressed. So using electricity slowly and using less of it at such times is encouraged. Some utilities already have tariffs like this in place.

Domestic microgeneration can also provide 'reverse dynamic demand' support by dropping off the grid (stopping generation) relatively quickly as the mains frequency (and voltage) increases, but hanging on grimly pumping in energy for much larger relative falls from nominal central values, ie by being deliberately asymmetric in its response to grid loading/stress.

UK Domestic Appliances with Dynamic Demand Potential

Here I try to summarise potentially dynamically-controllable domestic loads on short (very temporary, of the order of seconds) to long/permanent (days) timescales for the UK.

The UK grid already has demand/supply matching on these scales, though below ~1.5 hours is more typically considered 'balancing' and dealing with unplanned plant (and transmission) outages/failures, or even 'demand pickups' after popular TV and sporting events. The medium term of 'hours' is often concerned with 'load shifting' to lower peak demand and to reduce infrastructure and 'peaking plant' costs. The longer end of the scale is dealing with extended supply imbalances, such as a long series of dark cold calm days reducing renewable energy supply.

I attempt to tabulate the different appliances' possible responses by short/medium/long term timescales in the UK.

Short-term deferrable demand of a few seconds to a few minutes can help with a crisis/emergency such as the unplanned 'tripping' off-line of a major generator such as the 1GW+ of Sizewell in May 2008. Clearly almost every domestic heating element could stop drawing power for up to a few seconds without disaster (or probably anyone noticing), almost every fridge could avoid starting its compressor, every washing machine and dishwasher could stop spinning/pumping/etc, to give the grid a few seconds to breathe and other standby generation to start. For example, in the UK "a number of Open Cycle Gas Turbines are set to operate by low frequency relays (set at 49.5Hz and 49.6Hz) and should [reach] full MW output [] in around 5 minutes", so filling in that 5 minutes with dynamic demand control is potentially useful.

Medium-term deferrable load helps cut regular daily peaks (such as from everyone cooking when they get home after work), for example by delaying electrically heating water or running a heat-pump for central heating. Programmes/tariffs such as Economy 7 (less than 30% of all domestic supply) already provide incentives to shift load, including other loads such as dishwashers that can be put on timers. Increasingly dynamic ToD/HH/emergency pricing and dynamic demand might work well together to easily shift more load and trim peaks.

Long-term deferrable consumption would allow less standby generation to back-up intermittent renewables such as wind and sun and wave. Note that in National Grid's recent SYS (Seven-Year Statement) it assumes that if the UK increased wind name-plate capacity on the grid to ~25GW (somewhere between one third and one half of unrestricted winter peak, ~10x the 2008 installed capacity, a reasonable target over the next ten years or so, and by which time most domestic appliances would have been replaced and could have dynamic demand controls in the replacements) then the UK could only retire 5GW of conventional thermal 'backup' plant. (That is, retiring only 2GW more than if just 8GW of wind was on the grid.) Normal grid margin for non-wind plant is ~20%+, so the intermittency gap here is ~60%, and ideally, we'd like dynamic demand to allow the retiring of that extra 15GW ~10 years from now to eliminate most/all of the extra system and balancing costs related to wind.

(Note that in an extended energy shortage the National Grid can ask the Distribution Network Operators (DNOs) that connect up to each house / end user (with a Demand Control notice, only expected about once per year) to reduce demand from their users by lowering supply voltage (aka 'brown out') which cuts instantaneous consumption from plain-vanilla resistive loads ~10%, but doesn't have much positive effect on more complex and feedback-driven loads such as switched-mode power supplies and motors. Increased current-draw by non-resistive or regulated loads may in fact eventually increase distribution losses and even shorten equipment lifetime, but in conjunction with the long-term dynamic-demand mechanisms outlined in this document, could continue to suppress demand by useful percent.)

This table includes estimates for typical 2008 electricity demand levels and what might be deferrable on the given timescale, with examples of specific load components that are and are not deferrable.

The estimates of what fraction of mean load is deferrable in each timescale are largely educated guesses by me, with some justification given.

The implication of the numbers above is that dynamic demand might be able to perform almost all the 'extra' essential grid support on short timescales for 25GW of wind on the UK grid compared to 'normal' thermal plant instead of retaining less-efficient peaker and frequency-response backup, ie ~15GW+.

For dynamically deferring load a few hours to cope with medium-term dips in intermittent sources such as wind, and allowing highly-efficient thermal plant such as CCGT to step in, it seems as though grid-friendly appliances might obviate maybe 33% of the 'extra' grid support/backup otherwise needed, ie ~5GW+.

For long-term dips/outages of the order of days (such a long cold still snap) the dynamic demand mechanism may only be able to help suppress demand a little, ie maybe ~2GW+, but 'long-term' is not the primary focus of this technology, and for long time scales explicit information campaigns (etc) may be needed, plus the firing up of more baseload-like plant if available, and other System Operator tools such as voltage reduction and rationing.

When to Trigger Dynamic Demand Mechanisms

National Grid as the grid System Operator has various targets and legal obligations as to how much it is allowed to let parameters such as frequency wander from their nominal/central values.

The limits on the 50Hz-nominal UK grid frequency are:

• Operational +/- 0.2Hz
• Statutory +/- 0.5Hz (1%)
• Maximum drop from an Infrequent Infeed Loss Risk (1320MW) is -0.8Hz

To protect the majority of the grid, automated 'low frequency' (ie frequency-sensing) relays start shedding load to protect the grid at -1.2Hz (48.8Hz) with 5% of total grid load intended to be disconnected in this first tranche.

There will naturally be some 'wiggle' in the frequency as demand and supply will never balance exactly to the Watt at every instant, but if the wobble gets too big then something is going wrong and needs fixing.

Primarily I am interested in moments where the grid is short of power, ie not meeting demand, as if you started to put the brakes on a car or bike too hard slowing it down, and thus when the mains frequency drops below its nominal 50Hz value. I assume that too much power available is an easier problem to handle, probably by other mechanisms than dynamic demand, such as throttling fuel and turning turbines out of the wind. (Though consumption such as auto-defrost cycles for fridges could be delayed until there is excess power available without ending up using more overall.)

An analysis of figures supplied by National Grid for August 2008 (.xls) (a frequency sample every minute for the entire month) indicates only one sample out of nearly 45,000 potentially below the operational -0.2Hz target, but ~3% at or below -0.1Hz (ie frequency ≤49.9Hz) and ~17% at or below -0.05Hz (ie frequency ≤49.95Hz).

Thus if dynamic demand control in domestic appliances is to be of help with (and reduce the cost of) 'balancing' of today's UK grid, it would have to start to intervene before the -0.2Hz operational limit. Somewhere between one half and one quarter of that limit looks plausible, especially if the response can be graded and smoothed, where it will help a small proportion of the time but with the current/normal grid mechanisms doing the fine tuning.

An analysis of figures supplied by National Grid for 2008-05-27 (.xls) where a series of generator problems resulted in 500,000 people being 'load shed', ie getting a 40-minute power cut, suggests that over ten minutes were spent outside normal operational limits, and about nine of those outside statutory limits, with ~5% at or below -0.1Hz (ie frequency ≤49.9Hz) and ~21% at or below -0.05Hz (ie frequency ≤49.95Hz). For a large chunk of the day various DNOs (regional distributors) were helping curb demand by reducing voltage too.

This initial analysis suggests that, for example, all non-critical heating loads (for example) could be dropped and stay off when the mains falls below the lower statutory frequency limit (49.5Hz) since the grid is clearly in significant trouble at this point and this condition does not last long even on the bad day examined above, with all other significant adjustable activity briefly suspended or reduced (continuously/stochastically in proportion to the frequency deviation if possible) as the frequency drops below 49.9Hz--49.95Hz, with load restoration in particular randomly staggered by time and/or frequency to avoid introducing instability/oscillation to the grid.

Specific Suggestions for Domestic Appliances

This applies to all domestic appliances with a significant deferrable load, principally heating/cooling, including but not limited to:

• Fridges/freezers (cooling).
• Dishwashers and washing machines (water heating).
• Electric ovens, microwave ovens, cooking/hob rings.
• Electric kettles especially any rated above (say) 1kW.
• Electric immersion/storage heaters and heat-pumps/air-con.

These will usually have no way of directly detecting grid load/state themselves other than by monitoring frequency, and frequency-based response is thus assumed for them. This mechanism should work fine in conjunction with other peak-shifting schemes that may only supply power off-peak but which can cause a surge when large numbers of such off-peak devices are powered up simultaneously at the start of the off-peak period.

The basic principle is that at times of grid strain:

• heating elements are modulated to draw less instantaneous power
• thermostats are slightly adjusted so as their regulated systems use less total energy

The first mechanism primarily deals with short sudden drops in supply vs demand, whereas the second is also a longer-term load-trimming mechanism.

The power modulation scheme might work as follows:

• For the initial few minutes of a severe energy shortage (ie a large drop in mains frequency, by more than ~0.2Hz for the UK) heating/cooling would turn off entirely. (This off period might be limited to a few seconds for interactive/fast applications such as kettles, electric hob rings and microwave ovens.) Mechanical agitation in washing machines, dishwashers, etc, and circulation pumps for central heating, etc, would continue once started but any new activity might be deferred, as would be the case for heat-pump and fridge/freezer compressors. As a bonus of having this mechanism in place, having some appliances 'soft start' by deferring (full) power consumption when mains is first connected would also help the grid after blackouts and more routinely (for example) when off-peak circuits activate.
• If the grid does not fully recover within this initial short interval above (but is back within normal operational limits), some appliances such as dishwasher and washing machine could almost certainly turn back on heating elements but at (say) the minimum of 1kW or 50% of maximum rating (and in conjunction with reduced target water temperatures described below) without massively extending the wash cycle time; such appliances need never revert to full power when the grid is strained, except possibly upon a manual request via an 'urgent' momentary button. Modulatable heating/cooling such as immersion heaters could run at significantly (say 50% of maximum) reduced power for a longer period (a significant multiple of the initial completely-off period) without compromising the ability to store/shift enough heat through the day.
• Immediately in the case of mild grid strain, else after any heavier modulation above has run its course, and indefinitely to mildly reduce continuing demand, heating/cooling power could be down-modulated slightly, say by 25% to 10%. (Other variable-speed/load devices such as some circulation pumps and inverter-driven heat-pumps and a/c could run slightly below maximum.) Not enough to be a severe annoyance, but enough to reduce peak household demand and some waste. (Though longer running periods to achieve 'fixed' goals/targets are likely to result in more overlap from separate loads, and therefore less aggregate grid peak load reduction than otherwise apparent.) DNOs may also at this stage, under instruction from the grid operator, be reducing mains voltage (inducing a 'brown-out') to reduce demand (~10%) mainly from resistive loads (eg heating elements), though a large population of grid-friendly appliances should reduce the need for this (already rare) step.

Note that under any significant grid-strain, thermostats should be adjusted where possible (ie with an electronic rather than mechanical thermostat) for the entire duration of that strain/shortage to defer consumption in the short term and reduce energy consumption though insulative losses (etc) long term, which implies lowering target temperatures for heating and raising target temperatures for cooling. The change in temperature should not be enough to be damaging or annoying, and in the case of washing appliances could often be compensated for by lengthening the wash time while continuing mechanical agitation. (Biochemical activity generally follows a "Q10=2" rule, ie that a 10K (ie 10°C) drop in temperature halves the speed, so dropping the wash temperature 10K might require up to twice the time to achieve the same result. The power of the heating element can be reduced to achieve the new lower target over a similar fraction of the extended wash/agitation time, thus reducing instantaneous power and total energy consumed. The wash time can be suitably extended to allow for whatever the unheated inlet temperature is in extremis, ie foregoing any water heating at all if need be, which at say 10°C lower in winter might at most double wash time again for a given fixed, very much reduced, energy consumption.) There may be other goals/targets for appliances other than temperature which can also be trimmed at times of grid load without significant problem; eg rinse duration, non-final-spin speed, oven/fridge light level, etc, to assist with both short- and long- term demand reduction.

All modulation of demand should be somewhat randomised to avoid itself making the grid unstable with synchronised mass switch on/off of the 'dynamic demand' devices. The delay to reduce or drop loads/demand after a frequency drop should be a small random number of cycles (not too long a delay since a fast response is generally helpful) with the ability to cancel or further delay the load drop if frequency is detected to rise/recover during the delay. Switching back to full load after the grid recovers should be randomly spread over a much longer time for stability. The duration of each segment (fully off and reduced/modulated load) should also have a random spread around the mean value. When power/load is being down-modulated but not entirely turned off, ideally this accomplished by using a constant lower-rated element (etc), but alternatively a PWM (Pulse-Width Modulated) type scheme may be needed, probably synchronised with main cycle zero crossings, but with a scheme to avoid appliances becoming synchronised to one another. The PWM control might, for example, be run from a simple RC oscillator whose frequency is allowed to drift with ambient temperature. In any case the scheme should not be tightly tied to time since power on to avoid accidentally synchronising behaviour after a blackout for example.

Dynamic Demand and CO2 Emissions

It's clear that the kgCO2/kWh pollution of grid electricity is not constant, and includes large daily and seasonal variations based on demand, resource availability and supply mix, etc, etc. For example, if load were totally constant and covered by nuclear and hydro 'base load' generation, then we'd be practically zero-carbon. Given that daily/seasonal peaks over base load require fossil fuel generation to fill at least some of the gap at the moment, the non-constant demand implies more CO2 in the day and in winter.

Thus overall demand reduction is not the sole lever for reducing CO2 emissions (absolute or per unit of GDP or whatever); load-shifting and peak-shaving can also contribute, since I believe that electricity is probably at its dirtiest when all-hands-on-deck peak demand is being serviced necessarily including all available and least-efficient fossil-fuel plant, diluting the low-carbon sources such as nuclear and hydro. (In the UK circa winter 2008 intensity is thought to have peaked at ~600gCO2/kWh and thus nearly 50% worse than at carbon-intensity lows during the same day.)

Dynamic demand along with pricing mechanisms could also be used to lower the CO2 emissions of our electricity supply by trimming the dirty peaks. Thus the System Operator might aim to always operate the grid frequency a little below the nominal 50Hz at (CO2/kWH) peaks even when supply and demand are in balance, to reflect the carbon-cost and defer a small proportion of consumption.

In conjunction with smart metering, for example, it might be possible to charge a higher rate when the grid is under strain, and a penal rate if instantaneous use is (say) over 3kW or 1kW per household (which corresponds to running one or more major appliance(s)); conversely, a discounted rate might be given for instantaneous use under say 40W (1kWh/day) or 100W, with the discount enhanced further with the grid overloaded, to encourage people to turn off standby, use efficient lighting and turn it off when not needed, and to buy goods that implement DD and defer load automatically as needed. At 100W per UK household which might reasonably cover lighting and TV (say), total UK domestic electricity consumption would be under 3GW.

(There may be some scope for the System Operator and DNOs to apply some of the existing gentler demand-control facilities at times of high carbon intensity to blunt load peaks a little and reduce CO2 emissions, such as by reducing voltage very slightly. It might even prove possible to pilot such a scheme without rule changes.)