We all know if we use less electricity we save energy.

But the energy we're getting is result of burning (not necessarily the literal meaning) the fuel. Even if we don't use that energy that fuel is gone forever. (Except saving money on bills)

My questions

  1. Am I wrong in above statement? If yes then how does actually electricity production work so that we're able to save it?

  2. Do power plant (or associate agencies) constantly monitor demand and reduce the production of electricity on real time thus saving fuel (hence energy)?

  • 8
    $\begingroup$ There's a place north of me where they electrically pump water up a mountain at times of low demand and run it back down through turbines at times of high. There's river flow hydro too which they can scale within limits of agreed water levels, but the pump storage has several times the river's wattage capability - until of course it runs empty. $\endgroup$ Commented Dec 4, 2020 at 7:31
  • 11
    $\begingroup$ ELI5 level: When less electricity is used, they put less fuel in the generators. $\endgroup$ Commented Dec 4, 2020 at 18:54
  • 3
    $\begingroup$ When people say 'save power' they mean use less. Storage is a completely different animal, and for the most part, nonexistent. $\endgroup$
    – Mazura
    Commented Dec 5, 2020 at 2:31
  • 3
    $\begingroup$ This reminds me of a colleague who seriously argued that not taking an airplane flight is pointless because the plane would fly anyway, just without him on board. He completely missed the bigger demand/supply picture. $\endgroup$
    – Michael
    Commented Dec 6, 2020 at 8:00
  • 2
    $\begingroup$ I absolutely love these questions that look dumb at first glance, but provoke really fantastic detailed answers that demonstrate the real complexity underlying a simple question $\endgroup$
    – coagmano
    Commented Dec 7, 2020 at 2:56

4 Answers 4


Electricity production optimization is a very complex subject. It is also affected by many parameters , which I will try to outline below.

TL;DR: Demand is constantly monitored and supply is constantly adjusted

TL;DR 2: Lshaver's excellent post is a suggested reading after reading this, because it expands and explains what happens at timescales ranging from the micro-second to minutes timescales.

Changes in energy demand during the day

First of all, the energy demand during the day, week, month of year can change quite a lot. Below is a typical day in New England.

enter image description here

Fluctuation in power energy consumption a typical year in New England from a 2018 report.

Fluctuation in power energy consumption a typical year in New England

As you can see you can get changes in the range of 50% to 100%.

Energy sources mix

Electricity power plants come in many shapes and utilize different energy sources. Nowadays, the most usual types of energy are:

  • nuclear
  • fossil fuels
  • renewable energy (wind, solar).
  • Hydro. (I am reserving a special place for hydro)

For the first two there are finite quantities of fuel.

Renewable energy has started in the last twenty years to gain more traction. In the following graph you can see that, for electricity production, the combined renewable energy is now on par with electricity from coal and gas. I wanted to avoid including RES in this discussion, because of their stochastic nature of supply (however in the end - see final section - I couldn't avoid doing so).

enter image description here

Trend in electricity production

Difference between different power sources

Each type of electricity generation technology has different characteristics. I'll try to limit myself to the relevant for this discussion.

  • nuclear :
    • Usually high capacity installations near GW
    • very slow to change output.
  • fossil fuels:
    • cover installations across the range (kW to GW)
    • many different technologies (e.g. gas turbine, internal combustion, combustion etc)
    • usually high degree of efficiency means that there is a very slow rate of changing the output, and vice versa (otherwise some technologies would have died off).
  • Renewables:
    • usually many small generators
    • supply is stochastic and there is limited control
  • Hydro: I'm separating aside hydro from the other renewables, because of the following reasons i.e.:
    • it has high output installations, that can go from 0% to 100% in a manner of seconds and can be controlled at will.
    • it claimed the highest capacity installation among renewable energy sources(at least up until the last few years - however that is expected to change). (Actually, its one of the oldest technologies for producing rotational energy).
    • It can be used as a battery, for other types of energy: this is actually what - to my mind at least - separates hydro from the rest. You can use pumped storage systems, and for example when supply exceeds demand store energy. That energy can be later be reclaimed (there is significant loss). It can be done (and has been done) with nuclear, wind and solar.

How is power production optimized

Usually in a power grid there is (hopefully) the right mixture of different technology power plants. To optimize the system you have to:

  1. install as many as possible units with high conversion coefficient (which are slow to change their output), trying at least the minimum demand (I'll call them base units)
  2. install lower conversion coefficient units, which are quicker to "respond", and use them as reserve units.

So the idea is to have the base units (A) work 24/7, with weekly, or seasonal changes, and the reserve units (B) to accommodate for the hourly, or daily fluctuations.

Of course you can imagine that putting, the stochastic supply of renewable energy in the mix doesn't help the optimization problem.

Actually in some places there are restriction on renewable energy due to the supply's stochasticity. I live in an island with a small isolated grid system about 700[MW]. Renewable energy installed capacity has to be less than 30%, because the rest of the fossil fuel generators cannot cope fast enough with the changes in power production. Imagine, if at some point the wind energy dropped from 100MW to 0, in a manner of a few seconds.

Why output has to be monitored

To put it simply, in most cases (there are some notable exceptions like solar), in order to generate electricity there is - at some point - a rotating shaft with coils.

In a simplistic analogy, a conventional generator is like a fixed gear bicycle, that the rider does a specific rpm.

  • On a horizontal plane, the velocity will be fixed.
  • However, if there is added resistance (wind or uphill) the rider will need to either put more effort or slowly the speed will drop and maybe reach another equilibrium.
  • When, the bicycle goes downhill, then the rider will be forced to increase the rpms, or use some of the effort to maintain the same speed.

In the electricity production the rpms on the bicycle analogy are related to the frequency of the grid (50Hz ,or 60 Hz). The grid frequency has an important impact on some electrical equipment. So, another important parameters is maintaining the grid frequency.

For large grids that might not be so much of a problem, but for small isolated systems that can be very challenging. Actually, is some cases, the grid operator can use wind turbines for "backpedaling a fixed gear bicycle", in order to control the grid frequency.

  • $\begingroup$ Comments are not for extended discussion; this conversation has been moved to chat. $\endgroup$
    – hazzey
    Commented Dec 6, 2020 at 13:54

Timescales on the grid

Power demand fluctations can be broken into timescales from micro-seconds to decades. On the "decades" end of the scale, the power industry and utility regulators work together to plan and fund construction of power plants and the associated transmission and distribution infrastructure. When you turn your air conditioner off today it doesn't affect these long term processes, but if you (and enough of your neighbors) do this consistently, then there will be a measurably decline in cyclic demand, which will be factored into the planning process.

But even on the micro-second time-scale, turning off an air conditioner (or even just a light) does result in less fuel being burned now. To explain how, we need to build up the timescales from micro-seconds to seconds (beyond that, NMech did a good job explaining the hourly to annual time scales).

0.000001 - Inertial response at microsecond timescales

Fundamentally, generators are large pieces of spinning iron. This applies to coal, nuclear, natural gas, hydro, and wind generators. All of this spinning iron has inertia -- a fact around which the system is designed (in the case of solar power, grid-connected PV inverters are actually designed to emulate the behavior of spinning iron).

Inertia is "the resistance of any physical object to any change in its velocity" (Wikipedia). If you touch a huge spinning wheel, it will slow down -- but this change in speed will be imperceptible, and not instantaneous, due to the wheel's inertia.

On the grid, touching the wheel is akin to increasing the load -- turning on a light, toaster, hair dryer, etc. The inertia present on the grid causes a time delay between the change in load, and the change in speed.

This time delay is where the first level of grid control comes in: droop control.

0.001 - Droop control at millisecond timescales

Droop control is a proportional control on the "throttle" of each generator. Each generator on the grid is programmed to increase its speed by x% for every y% reduction in the frequency detected on the grid (and decrease speed when the frequency increases -- the proportion x:y is the same in both directions). This x:y proportion is the droop percent. The frequency, in turn, is directly proportional to the speed of each generator.

Droop control is like a string connecting the needle on a speedometer to the "gas pedal" of the generator -- as the speed decreases, the string pulls the pedal to give it more gas -- once the speed increases, the tension is released and less gas is supplied. This seems a bit backwards, because normally we think of using the throttle to control the speed of a car, but in this case we're talking about dozens or more cars all welded bumper-to-bumper, so one throttle can't actually do much. But if each car has the same exact speedometer/string/throttle set-up, we start to see how this could affect the speed control of the whole system.

This system has two shortcomings, though:

  1. What about when the load increases or decreases outside the range of droop control (i.e. all the generators are at full throttle and load is still increasing)?
  2. How is a constant frequency maintained?

This is where automatic generator control comes in.

1.0 - Automatic generator control at second timescales

The "throttle" functions differently for each type of generator. For natural gas, it's similar to a car -- you adjust the amount of fuel burned and the power output is directly adjusted. For coal and nuclear systems, it's more complicated, because the fuel is used to produce steam, and the steam then runs the generator. Droop control directly controls the "steam valve."

Automatic generator control (AGC) incorporates the diversity of generator types into a single algorithm which can be applied across the grid. Variables in AGC include:

  • Droop percent, as discussed above
  • Frequency setpoint, fixed at the grid level (60 Hz or 50 Hz, depending on the continent)
  • Area control error (ACE), a power output adjustment amount which depends on how much power the node where the generator is connected needs to import or export to other nodes

Where droop control is simple and straightforward to implement (at it's most basic, it can be done using passive electrical components), AGC requires more complicated, usually digital controls.

10 - Minute timescales and beyond

After AGC things get more complicated and less standardized across the industry. At the minute timescale and beyond, variables such as fuel prices, weather forecasts, emissions rates, and load forecasts can be included in control algorithms. Design of these algorithms vary by region and utility. The Wikipedia article on regional transmission organizations is a good starting point.

tl;dr - how does this save fuel when I switch off a light?

When you switch off a light, grid inertia immediately kicks in to increase the frequency (speed up). This triggers the droop control algorithm across the grid to throttle down generators. The generators which can physically respond fastest are programmed to do so -- these generators also tend to be those directly burning fossil fuels -- usually natural gas, but sometimes diesel.

Even in the case of coal, however, there will be decreased fuel use, but not until AGC kicks in. Droop control throttles down the steam valve directly, but as the pressure in the steam vessel responds to this change, less coal will need to be burned to maintain the set pressure.


Yes, they monitor demand and due to differences in how different power stations are controlled they can increase or reduce output to match demand.

Some power plants like nuclear run at full output as they are slow to change, but others like Dinorwig (in the UK) can go from standby to max output in 12 seconds and reduce to half quickly.

Danish researchers did a paper on controlling their grid as the wind power fluctuates - worth looking for and reading.

  • 3
    $\begingroup$ The search terms the op needs are probably base load, peak load and grid balancing. $\endgroup$
    – mart
    Commented Dec 4, 2020 at 9:46
  • $\begingroup$ Thanks for answer. Keep saving energy! $\endgroup$
    – ShivCK
    Commented Dec 4, 2020 at 10:39
  • $\begingroup$ 12 seconds might sound like it takes a while, but that’s from nothing to the equivalent of a swimming pool of water flowing every second. The engineering challenges to have a “tap” that turns on that fast and to that extent is significant! $\endgroup$
    – Tim
    Commented Dec 7, 2020 at 12:17
  • $\begingroup$ @Tim if you googled it, you could read that at full power the flow rate is 390 m^3 per second - sure that will fill your pool! $\endgroup$
    – Solar Mike
    Commented Dec 7, 2020 at 12:44
  • $\begingroup$ @SolarMike only just though - that’s filling a standard pool to around 1.5m deep on average. $\endgroup$
    – Tim
    Commented Dec 7, 2020 at 12:47

Peak Demand

One of the biggest issues in the electricity business is managing peak demand. This is the maximum amount of electricity needed at any one time. Typically this will be the middle of the day on the hottest days in the summer, as that combines household usage (everyone is awake), office usage (everyone is at work, or at least before the pandemic...) and air conditioning (one of the biggest electricity uses in many areas - heating in those same areas is typically a mix of electricity and fossil fuels, plus insulation helps more to keep heat in than to keep heat out, thanks to all the people and computers constantly adding heat inside buildings).

You don't want to simply run at peak levels all the time. That would waste energy and money. Instead, you run some plants all the time (e.g., nuclear, coal, oil - which, to varying degrees, are very slow to change their output levels) and some plants as needed (e.g., natural gas turbines are relatively good at fast startup for peak demand usage). Solar and wind are largely uncontrollable, though solar at least fits the general "daytime demand" pattern. Hydroelectric power is sort of in-between: It can be adjusted to a certain degree and if designed with a pumped reservoir can function as a large battery/reserve power source. But hydro also tends more, especially on big rivers, to be a use-it or lose-it power source - if the reservoir is full and the water keeps coming in, you either generate power with it or you let the water run away without generating any electricity.

The largest, most constant-level plants also tend to be the least expensive per kWh. Running the peak-only plants tends to cost quite a bit, plus those are almost exclusively fossil-fuel based. In addition, there is significant infrastructure cost to building additional peak-only plants, and that cost gets spread over relatively low production. As a result, utilities have a significant incentive to manage or reduce peak demand. This can take several forms:

  • Batteries (real electric batteries or hydro as described elsewhere) - store power at low demand times, release power at high demand times.
  • Encourage insulation, installation of efficient lighting and appliances, etc. to lower overall demand.
  • Install devices to cut off air conditioning, water heaters and similar large-but-not-critical power using devices during peak times, often for a monthly bill credit.
  • Charge each user for their individual peak demand, encouraging customers to manage their own demand levels. This is more common for industrial and large commercial users than residential users.


One way, when peak demand can't be managed well, to handle this problem is by reducing the voltage provided to customers. Otherwise known as a brownout. In the US, electricity is normally provided at 120V/240V to residential and small commercial users. But most equipment can easily handle 110V/220V or even a bit less than that, with no ill effects except somewhat dimmer lighting (depending on the lighting technology used) and longer times for heating (electric water heaters, resistance electric heat, electric dryers). In fact, while frequency is extremely tightly controlled, voltage can vary quite a bit over the course of a typical day.

If you turn on your dryer, an extra 30A @ 240V = 7.2 kW is now pulled out of the grid that wasn't needed a moment ago. But that is out of perhaps a Gigawatt = 1,000,000 kW of total power generated/used in your area. So your 7.2 kW is spread across thousands of other customers and makes only the tiniest difference in power available to others, reducing everyone's voltage a very tiny bit (calculation left as an exercise). Conversely, when you turn off your dryer, that increases the power available to everyone else. If enough people make changes at one time - e.g., all the offices turn on lights & computers at 8:00am, or all the houses start cooking dinner at 6:00pm, then the utility will adjust the input by turning on peaker plants.

It is actually quite a bit more complex as power is shared across large areas of Europe, US, etc. and utilities will often buy power from elsewhere, if it is available at the right price, rather than turn on more plants of their own.

Batteries to the Rescue

One new technology is to use very large battery packs to store electricity (ideally generated by solar or wind that produce power "when they feel like it", but the electricity can come from any source) for use when needed. The "when needed" is for peak demand but also to help control frequency. Frequency is critical for a bunch of reasons, and can get out of control quickly if there are large sudden changes in demand. Plus even the best of peaker plants take many seconds to go from idle to full production. A very large battery pack can impact the grid in a fraction of a second, keeping the power more stable (a general improvement to quality & reliability) and lowering the need for peaker plants or allowing for slower ramp-up to meet peak demand. See the Hornsdale Power Reserve for a good example of this technology.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.