The classic case of users managing their demands was at the JET fusion facility, where they could call on the grid for only some of their required power, but had to install a ginormous flywheel + generator to supply the rest of the load for the few seconds of their pulse.
JET’s power requirements during the plasma pulse are around 500 MW with peak in excess of 1000 MW. Because power draw from the main grid is limited to 575 MW, two large flywheel generators were constructed to provide this necessary power. Each 775-ton flywheel can spin up to 225 rpm and store 3.75 GJ, roughly the same amount of kinetic energy as a train weighing 5,000 tons traveling at 140 kilometres per hour (87 mph). Each flywheel uses 8.8 MW to spin up and can generate 400 MW (briefly)
I’m not an electrical engineer but I am involved in specifying some pretty large rotating equipment, up to 2MW – specifically, combustion air fans for industrial furnaces and hydrogen plants (steam methane reformers.) I’ve also worked (on the combustion side) on power stations.
I’d say one factor you haven’t considered is that the stall current of large electrical machines is enough to destroy them, and managing this is just as important as managing the load on the grid. It’s a condition of most of the fan motors I’m involved in that they must be started off load. That means a valve must be closed to prevent any airflow before they are started, and can only be opened once they are up to speed. There are also rules to be observed about max numbers of start attempts from cold. When starting from hot, typically only one restart is allowed (and a cooldown period may have to be observed.)
Small motors <5kW generally have direct-on-line starters, meaning they are simply connected straight to the 3 phase supply. Medium motors (8-300kW) often have star-delta starters, so that they start with a star connection (meaning they start with the windings seeing the voltage between one phase and neutral) and then once they are running at a certain speed they switch over to a delta connection (meaning the windings see the voltage between 2 phases.) The ratio of these two voltages is always sqrt(3) and is usually 230V for star connection / 400V for delta connection in Europe. Very large motors (typically those too large to use the 400V supply) go back to direct-on-line starting.
Typically any large plant will have more than one train, and the starting current of an individual train will be higher than the running current, but less than the running current of all trains running together. The starting surge will therefore be within the capabilities of the grid.
There will however be coordination with the grid at startup, as others have mentioned.
I know that a paper mill on start up would demand power from the grid and this was planned for by phoning the grid people.
Once started, the mill had sufficient power to run itself and supply excess to the grid.
The mill forgot to organise startup once and the area suffered a power cut. LOTS of hot words were said! Only down for 10 minutes but a bit embarrassing 🙂
And as for your spinning reserves not being capable, places like Dinorwig can go from zero to full output in under 16 seconds (edited from 12 – now based on the Dinorwig website) (1800 MW), it was designed to satisfy peak loads rapidly.
They have the grid operator on speed dial.
You don’t get that kind of power without having a relationship with the grid operators, and you’re going to coordinate any large loads with them.
This is seen in the Chornobyl accident and the miniseries Chernobyl, where they asked the grid operator to let them drop off the grid to prepare for their test, but the grid operator demanded they stay at half power late into the evening. This was responsible for several factors in the accident.
I realize that’s different, but only in + and – signs, not in the basic principle. It’s still a matter that everyone must coordinate via the grid operator.
In your residence, the operating paradigm is you get to draw any load you want and the utility has to find a way to deliver it or the grid goes down. That doesn’t scale to the sizes you’re talking about. You can’t just throw a half gigawatt load on the grid. That’s the same as a half gigawatt generating station or transmission line tripping.
From a grid operator’s perspective. The ability to cover a rapid change in demand is called Frequency Control and Ancillary Services(FCAS). This is provisioned on the basis of credible contingency. This is usually considered as a trip of the largest load, generating facility, or transmission line. One of the operator’s jobs is to ensure there is enough capacity in the system to meet this contingency.
At the operational level, the increase in load decreases the mechanical energy in the rotating machines, and the grid frequency slows down with them. Generator governors responds to the decrease in frequency by increasing the motive power. On a thermal plant, this is opening the steam valve, on hydro, it means opening the tap, on a gas turbine, increase fuel flow.
For very large power requirements, the plant operator will inform the power system operator, allowing them to provision enough FCAS to meet the load increase of startup, or load decrease in the event of a trip/sudden shutdown when the plant is up and running.
To add to what some have said about working with the power company, the relationship can be complex. I work for a power company somewhere in the US and we have several large customers that draw power in the range you mention.
At least here, it is not normal for the power provider to be the same as the power supplier. We do not produce our own power. We buy it from, “the grid,” which is another company that is tied into the national grid.
Not only are we concerned about the needs of a particular plant from our distribution network, we are concerned about having sufficient power into our network. Electrical energy in this quantity is difficult to store. Usually, it must be purchased live.
Given this, we monitor live load constantly. I have never asked but I would presume that some of our engineers could tell you when each of the plants on our network has shift changes and their load dips. Small or scheduled dips, or changes from network-insignificant loads (one house at a time) are either planned or come out in the wash. Aggregate network-wide changes like winter heating at night or summer air conditioning during the day take excessive planning and often cause headaches anyway.
A plant stopping causes loud alarms and usually phone calls or meetings to decide what to do next. There are many people with a lot of experience who know how to anticipate, communicate and mitigate situations like this.
Similarly, when a plant like this turns on, either as its original start or after some stop event, there is a great deal of communication and planning. We would be likely to have seen detailed plans and load specs well in advance in order to be able to adjust as necessary. If you’d like to turn on your AC during the day in the middle of the winter, we won’t pay much attention. If you’re running a GW plant, we know this (because we built you an expensive substation that can push a GW) and we’ll probably call you frequently to ask what you’re planning.
Back in the ’80s I worked at integrated steel mill (integrated: start with coal and ore and produce steel coils, bar etc.) They had a position known as “Load Dispatcher” whose job it was to monitor electrical usage and keep from exceeding demand limits. At that time the facility was billed for both energy used as well as maximum power used (e.g. demand) at any time during the billing period.
One of the facilities at the plant was an electric furnace that refined scrap and used a lot of electricity. It was designed to be shutdown mid-batch should there be a need to trim demand. The Load Dispatcher was in contact with the local utility and could manage the plant load if there was a need to do so.
I don’t know how much has changed since then, but I suspect big users still coordinate with the local utility to manage load vs. capacity. I would hope that some of that is now automated.