Power satisfaction bug?

[SpiderDeity]

Was testing nuclear on a test/freeplay world I set up for this sort of thing. I needed power draw to test the turbines so I set up beacon arrays like such:


I found that if I had more draw than I could produce, the satisfaction "maximum requirement" shot up massively. Below, I have less than my maximum power connected, satisfaction is 100%, need is 300MW, production max is 320MW.


If I attach more (just slightly above my max production), the "need" displays 1.7GW!!!:


And indeed the beacons reflect this MASSIVE lack of satisfaction


What am I missing here? Is this a bug maybe?

[foamy]

I wonder -- maybe it's because the beacons are trying to make up their entire power pool at once, and so are trying to draw more than their rated fixed power? I dynamically turn off beaconed sections of my plant via switch, and one of the things I've noticed in the power graphs is that, when a section turns on, there's a brief pulse of higher-than-normal draw before things restabilize.

[Nosferatu]

I noticed the same thing when I did my brown out tests.
It looks normal for all things that consume electricity actively.
Drain type consumption results in what you observed.

I ended up using stack inserters that move things around but that approach is probably not good for bigger tests..

[Yoyobuae]

What am I missing here? Is this a bug maybe?

Beacons specifically have a rather large electric energy buffer compared to the amount of energy per tick they use. Specifically they have a buffer around 5x larger than they use per tick. Here you can see the impact of that energy buffer (look how the Satisfaction bar progressively "drains"):

(video, right-click and hit play or go to: https://gfycat.com/defenselessdimdonkey)


Those buffers usually have no any impact, that is until Satisfaction drops bellow 100%. When Satisfaction drops bellow 100% the main impact on entities is that their electric energy buffers will no longer be full on each tick. If their electric buffer is not full then they can no longer work at full speed because they don't have enough energy available on buffer.

In fact, Factorio assumes that Satisfaction is equal to the amount of energy required to fill all input electric buffers on each tick.

Normally the energy demand is just the energy usage per tick and the math works. But beacons will bloat the energy demand number with their larger than needed energy buffers. Specifically when satisfaction drops each beacon will appear to consume 2.4MW instead of the normal 480kW.

Laser turrets are another big offender. Normally they have only a 24kW idle draw, but when satisfaction drops bellow 100% their electric buffers can bloat it up to 9.6MW:

(video, right-click and hit play or go to:: https://gfycat.com/leafyuniformgilamonster)


These specific entities have no doubt been the prime cause of many electric power death spirals. Laser turrets are particularly nasty and can easily turn just a minor drop in satisfaction into an all out power outage since they have top power usage priority.

I noticed the same thing when I did my brown out tests.
It looks normal for all things that consume electricity actively.
Drain type consumption results in what you observed.

I ended up using stack inserters that move things around but that approach is probably not good for bigger tests..

Radars are just fine to use as power drain for tests. They don't have the annoying and unexpected power satisfaction behavior that beacons have, because they don't have large unneeded buffers (only 6.7% larger, so Satisfaction may at most be be 6.7% larger than expected).

[SpiderDeity]

Awesome, thanks. Makes sense, of course the buffer charges faster than the base use, and I knew other things like lasers had buffers like this (I've seen their usage spike up when reconnected), I just didn't think that other things had buffers.

Much appreciated!