Friday facts #151 - The plans for 0.14

[Fatmice]

I want to have a chance to do an accidental Chernobyl when nuclear power gets added. Imagine that stuff blowing up if not cooled properly (risk/reward), and when it goes the radiation cloud can mutate the biters into whatever freakish mutations you can dream up that should be a pain to deal with

Reactors do not blow up in the way your mind is envisioning it. This is a common misconception/misunderstanding from the numerous amounts of disinformation out there.

In all the nuclear power games I've played, the challenge is to keep the water at the right temperature. Too low and the fission reaction can't sustain itself. Too high and there's a nuclear meltdown.
This is best left to the circuit network, and a power outage here should be potentially catastrophic. (Water temperature keeps rising until the plant explodes.)

Temperature doesn't play much of a role in the rate of fission. It's not like a normal chemical reaction. If anything hotter water would slow down the rate of fission.

Without getting into the weeds of nuclear physics, hotter water can slow down the nuclear reaction. However, it really depends on how the reactor is designed and the type of reactor. If the water is phase changing into a gas, more voids will form and less of the water will be in a phase where it can act as a good moderator and a heat carrier. So coolant temperature can effect the nuclear reaction rate, but again it depends on the reactor type and design.

[neurofish]

If you introduce a possibility to change recipe of assemblers by easy automatic means you will ruin the game. Think, why.

[TheUnknown007]

If you introduce a possibility to change recipe of assemblers by easy automatic means you will ruin the game. Think, why.

I thought why [that would ruin the game]. How, exactly, would this ruin the game?

[Zeblote]

If you introduce a possibility to change recipe of assemblers by easy automatic means you will ruin the game. Think, why.

I can't see how that would ruin the game. It might be completely useless though because you'd have to build something insanely complex to really make use of it.

[ssilk]

Some info from moderator sight.

There are hundreds of ideas around nuclear energy.

Take alone this search: search.php?keywords=nuclear&terms=all&a ... mit=Search
48 matches/threads so far.
And this thread viewtopic.php?f=9&t=5 Electric Energy
contains 96 matches with the word "nuclear". ( search.php?keywords=nuclear&t=5&sf=msgonly )

I know most of the threads and what continuously comes is this:

- Nuclear energy needs a somehow complex process to get the "fuel". Something which heats the water. Uranium processing. More ores. See also viewtopic.php?f=80&t=189 New Types of Resources (Gold and Diamond...) / New Ores

- More realistic kind of water heating: Changing the aggregate (Water -> Steam), such stuff.

- Water circulation: The water as medium for energy transport is not wasted in the steam engines. Instead it can be cooled and reused and you spare some energy. This kind of technique is useful in the time before nuclear power, too. I think to a turbine.

- To built this all up you need a very high advantage compared to steam/solar. Otherwise you won't built this. Why is there no need to do this? The current best idea above this is some entity, which needs more and more electric power, about 100-1000 time more, than a current "standard" factory. Impossible to built this with solar. The new technique - that needs so much power - spans from energy for the space platform or rocket to energy for underground mining (which makes new ores also much more interesting), complex ore processing and much more.

- Many want, that running a nuclear power plant is not very stable process (like steam and solar), cause they want to have explosions. Either explosions of the pipe network (dangerous handling such a hot fluid and pressure) or dangerous handling of the kettle with the nuclear stuff in it. Which has a high chance of pollution of course.

- Which leads to the last point (which is more or less my own idea): Running a nuclear power plant can be "optimized" / being made really safe by using circuits. It works of course. But to get out the last 10-20% you need to check things automatically, control pumps and valves, beeing alarmed and so on.

[Zeno]

Well I wrote up a nice piece explaining why 'Combinators are necessary for Nuclear' to achieve the stated goal of AVOIDING "Some kind of 'mine uranium -> put it into boiler' wouldn't be a proper use of the potential at all" but my internet apparently ate the thing when i hit "preview" (sigh) so I'll simply summarize that post by saying:
- Please realise that at the moment, there is no requirement for using combinators anywhere in your factory; Just like there's no requirement to use trains. It's just a choice at the moment... However, many players may be (and are) missing out on a critical aspect of Factorio automation, management, and logistics. (I know I am, I've NEVER used combinators in my 250+ hours) Nuclear seems like it would be so complex and unique that it would require an equally complex and unique solution for the puzzle it could present the player with. Combinators offer that by staying true to what Factorio is about; getting a machine to do the work for you. Especially paired with the soon-to-be updates for combinators.

I strongly suggest that any nuclear reactor gameplay should require constant management, management that could be automated by combinators.

This can by done by having the reactor constantly damage itself while it's active, and not be able to effect repair unless it's shut down... but would take even more damage when it's not being run within an optimal band of resource flow; If the reactor hits a "red zone" then it should take damage rapidly, until it implodes.
... Unless a combinator is standing on sentry duty, ready to hit the "Shut Down" button for you.

[Extended]

I think that relating the nuclear power management with the circuit network would be a really great idea, causing difficulties for non-programmer people though.

To my mind, the nuclear power should not emit any pollution but in disasters - it would be a regress comparing to solar panels otherwise. But sending the wastes thanks to rockets would make them useful.

[Requia]

A response to pretty much everyone everyone nuclear is not particularly complex or unique outside of the fuel processing and handling compared to anything else in the game. yes nuclear power plants require constant monitoring, so do coal gas and solar plants (if only because you need to know exactly when to turn on the coal because clouds are moving in). There is no call for new game mechanics, even the radiation isn't unique, a coal power plant produces more radioactive waste than a nuclear plant. Yes, meltdowns are a concern, but compare the worst nuclear accidents to the worst chemical accidents.

Actual suggestions: a complicated fuel processing and maybe reprocessing setup for actually making nuclear fuel. And get the fuel rations into the right order of magnitude compared to coal and oil.

I'm very excited about the nuclear power. I hope you will succeed in making it an interesting alternative with its own unique mechanics.

Some ideas that I can instantly think of:

• one should be careful with nuclear energy - if we overload steam/solar, everything works slower or doesn't work at all but that's it. Stressing nuclear generators could overload them so player would have to pay attention to have some 'reserves'. Or solve this via circuit network through backup sources.

Nuclear plants don't work that way, they'll run at 100% until the fuel is poisoned, problems actually come when throttling them *down* (which can poison fuel prematurely if the reactor isn't specifically designed for it). I've been hashing this out on a nuclear engineering forum (unrelated reasons) and the consensus seems to be that the run it at 100% all the time model is used almost universally, except in France because they have a 100% green grid so throttling simply has to be done. A better model would be for nuclear fuel to run out regardless of whether or not you actually used the electricity. This would mean that overbuilding wastes fuel, (actually running out of uranium shouldn't be a problem but you might need new rail or extra fuel processing).

[Zeno]

causing difficulties for non-programmer people though.

Tutorial map for combinators.

[SagaciousZed]

The Highlights of nuclear power could come from other parts of the process
- Enrichment
- Reprocessing

Reprocessing may introduce interesting logistical considerations because this is something different from other fuel sources, by forming a cycle. Instead of just burn and forget.

On a side node, I would really like to see you deal with all that coal ash....

[ssfsx17]

Uranium Power

First, let's talk about fundamental mechanics:

- Change of how the maximum power of a turbine is calculated. Perhaps a turbine could instead be specified to be capable of handling specific fluids. Vanilla steam engines would only handle boiling water, while there would be higher-tier turbines that can utilize the full energy of steam and other new fluids. Right now, higher-level turbines have to be simulated by either having higher effectivity - which means they can use lukewarm water better than a vanilla steam engine and must therefore use other mechanics to stop them from taking in normal fluids. Or they need to use up more fluid per tick.

- Fluid phase-change mechanics. For example, what if the definition of water also contained a definition of how much energy is needed for water-to-steam? Then, buildings/recipes that turn water to steam would be more flexible in how they work, rather than changing in a fixed amount.

- Heat exchanging as a core mechanic - including the possibility that a fluid could lose enough energy to switch down to a lower phase, without needing any recipes.

Now, let's talk about Uranium Power specifically:

- What if a material could have a dynamic amount of "enrichment" in it? Then you would drastically cut down on the number of different recipes that currently need to be defined in order to enrich uranium. Instead, you could allow the user to build anywhere between 1 and 1 million chemical plants / centrifuges, each with their own tradeoffs.

- With the fluid changes, players can be a lot more creative in how they build up their nuclear power facilities. Right now, the mod managed by Fatmice has to script a lot of stuff and be very inflexible in order to work around all the game engine limitations.

[AssaultRaven]

I'm going to disagree with a lot of the above and say that nuclear power shouldn't have a meltdown mechanic or other disaster/failure mode that is unique to nuclear power.

We don't model boiler explosions, and a whole lot more of those have happened than meltdowns.

I also point out that naturally occurring uranium can be used in reactors without needing enrichment, provided it is young enough, or just less than on Earth, it's it's younger. Alternatively, it could require more enrichment if it's older.

[Requia]

Uranium Power

- Change of how the maximum power of a turbine is calculated. Perhaps a turbine could instead be specified to be capable of handling specific fluids. Vanilla steam engines would only handle boiling water, while there would be higher-tier turbines that can utilize the full energy of steam and other new fluids. Right now, higher-level turbines have to be simulated by either having higher effectivity - which means they can use lukewarm water better than a vanilla steam engine and must therefore use other mechanics to stop them from taking in normal fluids. Or they need to use up more fluid per tick.

I quite like the idea of changing the general mechanics of power. Something I've been thinking is that right now there's no room to give in steam for better plants down the line because the efficiency is already super high. My thought was that the existing basic steam engine would be dropped to 25% efficiency (on top of boiler losses) but coal increased to 24MJ/unit (this should balance approximately since you'll need less coal for furnaces but more for power plants). Later tech developments (this would be more than simply 'here's a tier two furnace' as there would be an option for something like a coal (or oil) gas turbine that produces hot exhaust that can power steam turbines) would get the total efficiency back up to 50%. This keeps coal competitive for anybody who is perfectly happy pissing off every biter within a kilometer.

[IronCartographer]

I do like the idea of making nuclear plants less responsive - right now both solar and steam power will happily provide exactly as much power as you need at the moment, but if nuclear plants needed time to ramp up and down their power generation (by which I mean 'the amount of water they are boiling') they could become an excellent choice for late-game baseline power production without completely displacing steam/solar/accumulators as something to handle the spikes of production speed changing / laser turrets / etc.

Just wanted to say that I love the sound of this, especially the idea of accumulators averaging out the load on the nuclear generation.

[Solyx]

Idea for Nuclear. Different mechanic than what's been suggested so far.

My idea takes inspiration from the 'Thorium Molten Salt Reactor' craze going about (yes it's a good idea, but also yes there are some fundamental problems with it).

Basically U-233 fissions, neutron hits Thorium atoms, Thorium turns into Plutonium, then eventually fissions into Uranium-233 (or something like that). Fuel cycle requires U233 to get going, but it acts as a catalyst after that, in that the U233 is produced at a slightly higher rate than is consumed, and basically replaces/breeds itself.

This is the interesting part. You need a minimum amount of U233 to make the reactor work, but IRL breeding U233 is an incredibly slow process. We're talking years to double the amount you have. So scaling up such a thing takes a long time.

That's your mechanic - needing to breed fuel over a long time. Make nuclear so great and compact that everyone should WANT to move to it, but make scaling up nuclear a very slow, deliberate process they need to invest time in.

:::Example system::::

Add an Ore Processing Plant that consumes 10MW and takes in, say 50 stone ore and outputs 1 Thorium (say, 5GJ energy content, net 3.2GJ. 1 stack of coal = 400MJ for reference) after 180 seconds and also advanced a production-module-esque bar by, say 1%. When that bar gets full, a single U233 item also pops out. Being fed continuously, it would take 5 hours to generate that first U233 ore - can't short cut it either since you can't produce a single item double-fastwith multiple plants.

Add a nuclear power plant - it consumes 1 Thorium and 1 U233 over 100 seconds, producing 50MW. It also outputs a U233 (replacement), and advances a bar to generate an addition U233 by 10%. So this single plant will generate an addition U233 every 1000 seconds - or every ~16 minutes. So after 16 minutes you can start up another power plant, and then you can start up two reactors after another 16 minutes, and so on and so forth. Plus whatever all your Ore Processing Plants churn out. A 5GW power grid would take 100 plants, or about 7 quarter-hour-doublings, and net-consume 50 stone/second.

You could implement this all right now, with the only limitation being the need to output multiple items (or do manufacturing plants processing crude oil barrels already do that? maybe no coding-work needed.)

:::End example:::


You can futz with the numbers and everything to get the kind of production/progression curve you want, but that's the general gist. You make a Breeder Program. Long prep-time, slow-but-exponential buildup and payoff. It requires the player to make a large investment and aggressively mine and transport stone (50 stone/sec will require mining a lot of different average-sized stone patches simultaneously). It rewards players for stockpiling the mountains of Thorium they're going to generate before they can start using it, which also messes a bit with the min/maxers in that a specific ratio of power plants to ore-processors doesn't make as much sense. It provides a very large, slow-burning, compact power source. And once the player establishes their nuclear industry, they'll eventually have more U233 than they know what to do with. Now they can plop down a few power plants and a box of thorium anywhere without giving it a second thought, and power outposts indefinitely. Post-energy-scarcity world.



Anyway, that's the idea. Anyone have any refinements to add?

Edit - just thought of my own refinement. Coal becomes mostly worthless once you have nuclear, and it's far more abundant than stone. Use coal in the ore processing plants instead - makes everything cleaner.
Edit 2 - Add nuclear trains, and other such things. Takes 1 U233 and runs on Thorium. This gives a reason to keep up with collecting U233, and gives more advantages to a Nuclear Plant. No more train refueling etc. Single thorium would be equal to 12.5 stacks of coal. Stack of Thorium would be like 500 stacks of coal. Screw having to manage refueling a complex train network ever again. You can make this a general trend - the nuclear plants are the initial goal, but the U233 breeding infrastructure is the real prize.

[Furan]

But don't forget that it's a game not nuclear power plant science class. A healthy mix between realism and toy is mandatory.

nonononono, "nuclear power plant science class" is exactly what we need.

Just wanted to say that I love the sound of this, especially the idea of accumulators averaging out the load on the nuclear generation.

I was thinking more in terms of dynamic switch-off and loaders to overcome surges and stabilize production. Technology research queue would be nice.

[Zulan]

Glad to see that parallelization is finally being tackled. I do however believe the approach is unnecessarily complicated with a couple of significant issues:

• 9 divides very badly by common core counts
• Load balance for large patterns in which an inserter is always placed on specific bins
• This has to be reflected by memory management. Let's say (1)-inserters and (2)-inserters end up on one cache line, then performance is going to be terrible.

In principle, Factorio should be great for parallelization, most expensive computation is with respect to a position in a discrete 2D grid. Entities have locally limited interactions.

Now the FF implies that Factorio writes and reads the same local state during one tick so that the order of these updates matters. This could also explain things like the dreaded upwards inserter being slower than others. I wonder if it would be too much of a stretch regarding memory usage to use a double buffer for the mutable state. Then parallelization becomes almost trivial. Just assign target areas to threads, add halos of maximum reach around the boarders. So a few entities might be computed twice, but their output is discarded if not applied to a responsible position. I suspect the tiling for parallelization should be larger than 32x32 chunks to reduce the halo overhead. In addition some logic should be greatly simplified and odd effects like the north-inserter could vanish.

Of course this analysis involves a lot of guessing about the simulation, but I hope the devs consider established parallelization techniques from scientific simulations in Cartesian space.

Oh and global stuff such as power & production graphs really shouldn't be any issue. Just aggregate the date at the end of all each local updates

[Tev]

Nuclear power. Circuit network launching rockets. Multithreading!

Such an awesome blogpost.

I love you guys <3

[roidal]

Some kind of "mine uranium -> put it into boiler" wouldn't be a proper use of the potential at all.

In my opinion nuclear power would only be an advantage if the power/size ratio is much higher than that of actual thermal powerplants.

Maybe this should come with some risk. For example as long as uran/energy is inside the powerplant you have to provide enough water for cooling (one offshore-pump?).
If cooling fails (pipes get destroyed) the powerplant overheats and explode destroying everything around it.

[TheKingOfFailure]

My 2 cents:

-We should definitely be encouraged to build the nuclear reactor far away from our base so we have to transport things very far to get to it.(ie. radiation can damage factory/make it inhabitable)
-The power plant itself should be a large multi-building system, making it highly customizeable(more pollution/power/efficiency depending on your setup)
-Possibly make it so certain processes can't be done close to each other to spread out design/ logistics challenge
-Bots cant be used on site of the reactor because their circuits get messed up. This forces the use of spaghetti belts
-Must have a way to handle waste, from using it in weapons to just taking it by train very far away and dumping it on the biters' doorstep
-Anything handling nuclear materials must be specially designed(at least containers/wagons) or they will be slowly damaged, eventually being destroyed and dumping their radiation into the atmosphere

Then again, I'm no nuclear engineer, just your average Factorio Joe