Will AC power distribution become obsolete?

[Someone]

Distributed control means by definition that there is a central coordinator that distributes its control actions across the system.

This is why you have the reference to co-ordination used by the Wiki article, and by yourself.

You cannot use the word co-ordination and then imply that everything acts independently without reference to some central or out-of-band (Wikipedia words) signal.

It's a plain fact. If you have lots of individual generators all acting independently without reference to the other actors, you will not have frequency control, and you will not have grid stability.

This already exists, generators that sync up to the grid and have zero outside communication. PV solar for example. Completely distributed control.

Any number of generators with simple droop cycle control can be connected together, disconnected, joined, stopped, started, in any arbitrary order and network size. Without a single central authority or control.
https://en.wikipedia.org/wiki/Droop_speed_control
There is no central leader or actor required. Droop control still achieves frequency stability and control without any "central reference". That method is how the grid has been historically controlled, and, still is the primary method. Droop control has worked before we had crystal oscillators and high speed control systems that only make it easier and better.

Your definition of distributed control is the complete opposite of conventional usage:
https://en.wikipedia.org/wiki/Distributed_control_system

A distributed control system (DCS) is a computerized control system for a process or plant usually with many control loops, in which autonomous controllers are distributed throughout the system, but there is no central operator supervisory control.

[gf]

Inertia of the grid frequency does not require energy storage. Simply being able to regulate a plant's power output up and down provides frequency inertia.

Even the primary reserve provided by conventional power plants is "virtual inerta" at the end, since it is provided via a local control loop which adjusts the supply of steam, water, etc. Only the response time of the control loop must be bridged by the inerta of the rotating mass (momentary reserve), since the output of a conventional power plant cannot be ramped up and down at unlimited speed. Power plants which cannot ramp up/down fast enough are not eligible to participate in the provision of primary reserve.

Distributed control means by definition that there is a central coordinator that distributes its control actions across the system.

As long as you allow some frequency and/or voltage droop, you can build grids without a central coordinator, and without any communication besides the grid itself.
Here is an overview of control methods for grid forming inverters: https://www.mdpi.com/1996-1073/13/10/2589

Today's public power grids use a combination of centralized and decentralized control. Primary reserves are activated by decentralized control (fast automatic reaction), but anything beyond that is still controlled/dispatched by the grid operator (slower). Note that there are also commercial and political considerations which may conflict with technical objectives. Who is willing to voluntarily reduce production and renounce money? Let the others do it. But they don't want either. And there are policies in place that give priority to renewable energies and do not allow renewable energy sources to be throttled (until the collapse of the grid is imminent).

There is one widespread weakness that decentralized control based on frequency droop still can't solve, and that is local grid congestion, because the frequency can't be different in different regions of the grid. It could be possibly solved by using voltage droop as additional control criterion.

[NiHaoMike]

How about require all new residential resistive heating devices above 1kW (except kitchen appliances and tools) must have a built in device to throttle consumption to 1kW or less during high demand times? (Specifically targeting clothes dryers, water heaters, and permanently installed heating, note that heat pump versions of those are not included as as they are not resistive heaters.) It could be done with a microcontroller measuring the grid frequency taking the average to self calibrate so no high accuracy frequency reference is needed.

[madires]

Over here we already have a simple load control for heat pumps and a new regulation for limiting power consumption of heat pumps and wall boxes. The old one is a simple on/off contactor remotely controlled by the local grid operator (benefit: kWh slightly cheaper). A law defines how often and how long the heat pump may be shut down (max 6 h per day). A new regulation allows the local grid operator to limit heat pumps and wall boxes to 4.2 kW to keep the grid stable (if supported by the devices).

[paulca]

Distributed control means by definition that there is a central coordinator that distributes its control actions across the system.

That would be a central controller or master controller pattern. 

Distributed control by definition means that "control" is distributed.  This means you have valves, switches, process controllers which are "active".  When a temperature sensor sees a value is low, it directly asks the heater for heat.  It does not go back to mummy for permission that will have been prior arranged in the systems modelling and setup.

Consider home heating control systems these days.  In many cases the radiator sensor/valve is the entity which directly asks the boiler to fire if it needs it.  Personally I do not implement mine like this.  I use controller / controlled pattern, although it's decoupled as "best effort", serve and forget.

The thing about any distributed system it's difficult to synchronise things.  For some IT specific reasons, electrical/optical reasons and just plain old speed of light reasons, causality and stuff.  As a pretty rudimentary example, consider that while you can know the "round trip" delay between 2 nodes, it's is very close to impossible to know the one way trip.  Not when you are in the micro-second latency across continents.  To gain proper sync between nodes you need to use NTP which is a distributed time synchronisation protocol.  However it still won't get you past relativity.

The hardest part is keeping everything "inter-operatable" - may happen concurrently, distributed and not conflict.  Where that is not possibly and distributed entities may begin conflicting or competing operations that these are dealt with in a sensible fashion, aborting, rolling back, prioritisation, but first you have to detect conflicts and deadlocks.   The more complex the system, the more nodes and the more line crosses in the diagrams and the more difficult it is to get right and usually the bigger mess when it goes wrong.

For the leading edge of "distrbuted control systems", you want to look into the world of "Fleet" smart cars.  At the leading edge of that they are hoping to begin controlling cars, or rather having cars control themselves based on information shared with other cars around or ahead.  If you allow this to remain enabled then your highway cruise control can adapt to keep you "in the flow" and not bunched up with clumps of traffic waves and slowing you down etc.  If an accident is reported 5 miles up the road it can notify you via the nav that a divert is prudant. etc. etc.

[IanB]

Distributed control by definition means that "control" is distributed.  This means you have valves, switches, process controllers which are "active".  When a temperature sensor sees a value is low, it directly asks the heater for heat.  It does not go back to mummy for permission that will have been prior arranged in the systems modelling and setup.

Yes, I know this. It was late and my brain was foggy. I was thinking more about multivariable control where control actions are computed allowing for interactions between them.

With a typical plant DCS, two things are usually true: Firstly that each control loop (flow, pressure, temperature, etc.) is notionally independent of the other loops, so that each control loop can act autonomously. Secondly, that there is a central control room, where operators will monitor everything and take human action to adjust things like set points if needed.

A distinction on a power grid is that all the generators or suppliers are not independent of each other. They are all sharing responsibility for feeding power into the grid. If one should independently increase power output, another should decrease to keep things balanced.

This feeds into all the other things you mentioned, which when summarized means there is significant complexity in the control problem, and operating power grids is difficult.

Thanks for the comments.

[Someone]

Inertia of the grid frequency does not require energy storage. Simply being able to regulate a plant's power output up and down provides frequency inertia.

Even the primary reserve provided by conventional power plants is "virtual inerta" at the end, since it is provided via a local control loop which adjusts the supply of steam, water, etc. Only the response time of the control loop must be bridged by the inerta of the rotating mass (momentary reserve), since the output of a conventional power plant cannot be ramped up and down at unlimited speed. Power plants which cannot ramp up/down fast enough are not eligible to participate in the provision of primary reserve.

Different organisations split the distinction between inertia and frequency control at different periods/rates, so there is no direct equivalence but I agree they are the same action just at different time scales.

Existing synchronous plant has the inertia in rotation mass which is sub cycle response time, inverters are great at providing that (or even active filtering of harmonics). Inverters could be programmed to look and work just like a coal plant with the same dynamics, or they could be used to produce new and more useful control paradigms.

[Siwastaja]

In terms of inertia, PV inverters which ramp down power output on overfrequency are completely equivalent to classic hydro or coal plants which also just ramp down power output on overfrequency. These plants cannot sink power (and store it), either; modern stored hydro plants can, and that is more equivalent to a battery hybrid inverter.

Therefore, "energy storage" is a red herring. Responsivity is the correct metric.

From stability viewpoint, if everyone (or significant part of) producers can cut down production, this is enough, because there are always consumers.

This is equivalent to one-quadrant DC lab power supply having good transient response against overvoltages with load resistor attached to it. Only when load is zero, you need to add sinking capability.

Energy storage (implying bidirectionality) and central controls can improve things further, but as Someone says, are not mandatory basic parts of how grid operates, and were not and will be not strictly necessary, as grid-tie inverters already do what existing production does.

[tautech]

One must be careful with generalisations.....

In NZ we have one particularly elaborate hydro system based on generation from Lake Taupo (616km²) headwaters and its outfall with 9 further hydro dams totalling 23% of NZ's current energy requirement.

Add to this a operating range of Lake Taupo's level of 1.4m variance which is both used as a battery to provide continuity of generation through drier periods, be they summer dry times or winter when precipitation falls as snow only to be released in the spring/summer months.

[coppice]

One must be careful with generalisations.....

In NZ we have one particularly elaborate hydro system based on generation from Lake Taupo (616km²) headwaters and its outfall with 9 further hydro dams totalling 23% of NZ's current energy requirement.

Add to this a operating range of Lake Taupo's level of 1.4m variance which is both used as a battery to provide continuity of generation through drier periods, be they summer dry times or winter when precipitation falls as snow only to be released in the spring/summer months.

Mountains make a high population density difficult to achieve. Mountains near the ocean make hydro power work well, and a low population density consuming that power makes a sweet combination.