The perfect OCD compliant nuclear power plant! --- Now v2.4!

[mrvn]

The problem is what you already stated. While nuclear reactors heat up at a constant rate, the heat has to reach the heat exchangers and this is quite slow. So you cant run them on-demand very well. If you store some of the power in steam you have a good buffer until the rector is running at full power again. I think 3-6 steam tanks are a good number per effective reactor.

As you say heat takes time to travel along the heat pipes. In my setup the reactor gets to ~800°C and the last heat exchanger never gets above 500°C. So I wonder if it wouldn't be enough to have a steam tank on the last heat exchanger. It will be the first to stop producing steam and as soon as it drops you throw in some more fuel. If the heat pipes + exchangers are already sufficient to store a fuel cell worth of heat then no steam tank should be needed other than as detector and to smooth out spikes.

I think you made a mistake in your math though. or rather there is a faulty assumption.

This is because every heat exchanger and heat pipe can store 0.5 GJ. [ (12 pipes + 4 heat exchanger) * 0.5 GJ = 8 GJ ]

I assume this 0.5 GJ is the difference between 15°C and 1000°C. But the heat exchangers will produce steam as long as they are above 500°C. So when the steam in the tank drops you will still be at 500°C. That would mean you have only half the heat capacity free, or 4GJ. And if suddenly no steam is used up anymore you would have to store the other 4GJ as steam. Should an extra 16 heat pipes be added to catch those extra 4GJ instead?

[kann_]

My calculation already assumes that. The first 485°C are not considered in my calculation.
500°C to 1000°C = 0.5 GJ in heat pipes and exchangers (5GJ in the reactor according to the wiki).

Since you need 4 heat exchangers per effective reactor (1 Reactor = 1 effective Reactor, 2 R = 4 e.R., 4 R = 12 e.R.,...) there is always enough heat capacity.

I think you could add more heat pipes to increase the heat capacity without problems. I think it might be useful to add some heat pipes close to the reactor to keep it below 1000°C in extreme cases, but this might increase the delay of the reactor. I think you just need to try.

The idea to separate the coldest heat exchanger and measure the output with a tank should work. At the same times I think it is not worth it, because it will make the setup more complicated. As I said before, I think 3-6 tanks per effective reactor for intermediate storage are an easy and good solution to the same problem.

My setup is reactor - heat pipes/exchangers - steam turbines - tanks and I trigger a new reactor cycle if steam is below ~80% capacity.

[Laogeodritt]

The problem is what you already stated. While nuclear reactors heat up at a constant rate, the heat has to reach the heat exchangers and this is quite slow. So you cant run them on-demand very well. If you store some of the power in steam you have a good buffer until the rector is running at full power again. I think 3-6 steam tanks are a good number per effective reactor.

TLDR: If you have a very low or almost 100% consumption you could be fine with just one tank for measuring the load, but with unstable consumption or intermediate values steam tanks will help to balance things out.

The interesting part, which some reactor setups seem to get wrong is that the reactor should be restarted once the steam levels are dropping and not once the steam levels are already low. Dropping steam levels indicate that the internal reactor heat is getting low, which means the internal heat capacity is low enough to get refilled. While low steam levels show that both, the internal heat capacity and the steam tanks are empty, which might be too late to restart the reactor in time.

It's unfortunate that we can't measure the temperature directly via circuits; heat pipe, reactor or heat exchanger (input) would work, just need to get temperature somewhere in the system. That way we'd be able to move the entire control system into the temperature domain rather than steam, with pretty much the same control logic, and directly manage fuel cell insertion when temperature is low enough (at some point in the heat system, we can determine that empirically) that we know we can buffer an entire fuel cell. We could use the heat buffer in the heat pipes instead of a steam buffer, reducing area significantly, and avoid this "steam buffer is dropping" detection...

Just seems so much more elegant. Alas, no temperature measurement yet.

As for detecting when steam is dropping... I guess this could be achieved most compactly with a clock circuit and a D latch? Every clock signal, check if current value < stored value, and on that same tick store the current value to the latch. (I was going to suggest a long delay line at first, but then I realised that was stupid and I already knew how to make a better circuit. Answered my own question, as it were!)

As you say heat takes time to travel along the heat pipes. In my setup the reactor gets to ~800°C and the last heat exchanger never gets above 500°C. So I wonder if it wouldn't be enough to have a steam tank on the last heat exchanger. It will be the first to stop producing steam and as soon as it drops you throw in some more fuel. If the heat pipes + exchangers are already sufficient to store a fuel cell worth of heat then no steam tank should be needed other than as detector and to smooth out spikes.

This seems to make sense - my thought process going through this: at max electrical load, when the nuclear reactor stops the continued heat flow into the heat exchangers should cut off the last exchanger nearly immediately, as desired. At less than max load... if the load has been constant for a long while we can assume a temperature equilibrium was reached in the heat pipe (the last one would probably be higher than 500°C, because in this case we wouldn't have steam tanks that allow the heat exchangers to never get backed up - they would back up and intermittently stop drawing heat), and the continued draw of heat from the heat exchangers would steadily reduce the temperature until the last one hits 500°C.

A concern I had is that the last heat exchanger, without a tank, might back up more often and be producing less when the load isn't 100%, because it'd be farther along the steam pipe going to the turbines (last heat exchanger would be outputting into a higher-pressure pipe), so this might trigger the detection too early. However, since the detector tank is a good steam sink, and the output pump would not only prevent backflow but thanks to its repressurisation probably prioritise the last heat exchanger, I think this isn't an issue?

re: spikes, do we know how much delay/bandwidth limitation there is from heat exchanger to turbines? Mostly fluid flow limitations I imagine? Wondering how much steam storage would be needed per W of spike (I wonder if I can model this as an electrical voltage regulator circuit and start using some power supply design techniques...)

[kann_]

As for detecting when steam is dropping... I guess this could be achieved most compactly with a clock circuit and a D latch? Every clock signal, check if current value < stored value, and on that same tick store the current value to the latch.

You might get away with checking just the steam level instead of the actual drop. Start Reactor if Steam < 80%.
I might have overlooked something, but I think that should be sufficient.

re: spikes, do we know how much delay/bandwidth limitation there is from heat exchanger to turbines? Mostly fluid flow limitations I imagine? Wondering how much steam storage would be needed per W of spike (I wonder if I can model this as an electrical voltage regulator circuit and start using some power supply design techniques...)

From my feeling the delay comes mainly from the heat pipes. The delay in the steam flow is very low, especially if you use a setup with pumps: pipe-pump-pipe-pump...
Once the reactor is without fuel all the heat pipes balance out their temperature. If the reactor turn on again it has to build up a heat gradient so the heat can flow. Once it is build up the setup has full throughput. Meanwhile the steam storage needs to supply the turbines. This can be a bit tricky, because tanks connected to tanks have super low throughput. The solution I found was to pump the steam back from the tanks into the turbines. basically a row of tanks going away from the turbines and a row of pumps pumping it back to the first tank. The pumps are only active of the first tank is below 5k steam.

[mrvn]

As for detecting when steam is dropping... I guess this could be achieved most compactly with a clock circuit and a D latch? Every clock signal, check if current value < stored value, and on that same tick store the current value to the latch.

You might get away with checking just the steam level instead of the actual drop. Start Reactor if Steam < 80%.
I might have overlooked something, but I think that should be sufficient.

When you run anywhere close to 100% the steam will drop on every spike. Fluid dynamics mean that there is a delay between steam getting used and new steam flowing in. It basically needs to accelerate. Also you want to ensure there is some space for extra steam in the tank before you drop another fuel cell, just in case your heat pipes can't hold all that new energy. As mentioned on full use my reactor was around 800°C and the last heat exchanger at 500°C. So when fuel runs out it would stop producing steam fast. Way before the reactor has cooled down to 500°C. So, going by the heat calculations in this thread a full cell will produce more heat than can be stored. So some has to go into steam or be used up.You have to factor that in.

For me this isn't a problem since I didn't put down any solar cells, having the reactor as main power source with some steam left over from bootstrapping. I only have to buffer the difference between heat produced and heat consumed for one fuel cell. The closer you get to 100% usage the less you need to buffer.

re: spikes, do we know how much delay/bandwidth limitation there is from heat exchanger to turbines? Mostly fluid flow limitations I imagine? Wondering how much steam storage would be needed per W of spike (I wonder if I can model this as an electrical voltage regulator circuit and start using some power supply design techniques...)

From my feeling the delay comes mainly from the heat pipes. The delay in the steam flow is very low, especially if you use a setup with pumps: pipe-pump-pipe-pump...
Once the reactor is without fuel all the heat pipes balance out their temperature. If the reactor turn on again it has to build up a heat gradient so the heat can flow. Once it is build up the setup has full throughput. Meanwhile the steam storage needs to supply the turbines. This can be a bit tricky, because tanks connected to tanks have super low throughput. The solution I found was to pump the steam back from the tanks into the turbines. basically a row of tanks going away from the turbines and a row of pumps pumping it back to the first tank. The pumps are only active of the first tank is below 5k steam.

Don't forget that heat exchangers, pipes and steam turbines have an internal storage. Pipes can probably ignored. But steam turbines buffer spikes already.smoothing the steam flow into them.

[Chaoticus]

Great work, this is a very nice design. I've been testing it in creative mode and I made some minor adjustments you guys might want to use:

• Added green indicator lights for each reactor
• Added red/green indicator lights for the backup power grid
• Removed wiring to all but 1 accumulator (Reading the charge level from 1 is effectively the same, as far as I know)

Note that the red signal constant combinator has been removed.

Blueprint

Code: Select all

0eNrlXNuO2zYQ/RWtnu1AvEoy2gIpChR5CRZ56UMRGLLNtYXoBl02XQT+91LyZRWZtDhy5Y3ShbGwJPKInDkczgxJf7NXUSWyPExKe/HNDtdpUtiLv7/ZRbhNgqi+V75kwl7YYSlie2YnQVxfFWkU5PMsSERk72d2mGzEP/YC7We9NYP1uoqrKCjTvFUT7z/PbJGUYRmKQwOai5dlUsUrkUvo11dXq6IMyjBNJGiWFmHzVb6uhiHv2Mx+sRdz9x2T6LI7iVjXBYq6BKr/bXMhkvYrwo298OqWt65dtP+837duntqBlSK4aMgc40M7mAqEGIIg/woINQXhV0DYGWQj1uFG5PN1Gq/CpFGPAutVvlK6mzA/CNde8EbWZZ5Gy5XYBc+hrC7rHEGX8tkmPKvhKcyLcnnBk+cwL6sgapGsKTF/bx80KbVesxQ59VWcBXnTyIX9i6yRVmVWATAfD5jZi2xalZTLpzyNl2EiMezFUxAVYg9hj5TJOszXVVg2l7hLps41ZpJcMxub4um4yE0ZQK4wwDUaWHPvpHiuHVi52PQNK07rcW7W59lVJKKCVkrJM+qgP6h/Xc1ypu0f6iEFYgosZYd8pS296BHC7S4pcOqBZARE+4CQIRDvA8KGQF4fEDEDwk4fEDUDmqMzEtEgvZraLE+3eRDHwSoS8yITwReh6uR51Kpt62m0nG3rANPKvzetXcv6W/P4+J76dixKkTcj4gCwfA6k+7AMi2UWluvd0XbWs3pR5lUsRXDgunxvWooDz/n+lolZVv9cA3zfmiwKXlbB+svyOY2quoPOO+o7Lj1/JEPPZbZRugqiSMpWtlE2Vn5Pvy6zNHrJdmlyvL2v74v8ots7WbR5cKrd3EmTZRxkp1u1F3XUQdNmjSeUVOtISLOdi2B99IYOr4xFUQTbusinwyOLW5tUFJYUoiUJICyRpNV2Zz1VInqwPiSFrGXFQVLV3XqwlfTjUPrR8elH34B+GES/C6uMXEP+MeS1Pv60+Udv558L5R8bn3/4DfhHYfzjXf551/0k4ih8RxNz+f3ftOmKb6erBwqMTjEWbZzHVlxER4yLeidveFhUu7j/ZWCELrxc/5Kex0hI4V5jT+MCIx+mHvcN1EMnoJ4eY4IdkLZ8Xc7EgWnLewNt4QloCzR2CNZpA8FyPvjoCOJ7quPxx1cH8a7LX5ePwNAUIlaiENPA9JzZwOrAFBuHuG4fEjNFYm0kNY2g6n1v35Ky06Z7oTk2tbJccBYBaUTcymbF0n2ZR0Gc6f0RMqLzHCQv5S5MtgY+dFUI+aoorZ3Co7domGDrzlBYm1/rOr5a8+ebSdD9WSToG0sQmUmQOIN8Y6Qd6vfwhn8dMGMcpDSqP9yVedcDo1g75ytfoPWYCRrkMY+sNDoFpV2G1B0dEqCOdH4ywWamyftJTBN2jE0TMTRNZFCkMTLL8SRYjmAs1sYXhBqx+BxXoMnTGF1f/iX6RUGiW+MjDCREZ+oyJPqFYfmoMycyQ2PAYYEWUoK44DjLUbvuxGwhuhVmObA9LLjHl3B75IhMV9SJD47zNDKhDiy8UqqIImh0pWsNHuLYOvfMi7AJ5EV63CVKIGks5mhISMkQh/auyiI/vrIu44uOsihIWbo0CqVD/LK7KgtNQFmwkaObFykblAAe2Ul+nIST7PV5xTqHjnKoH6HJAFIXmq7VAXnQWVwH5EPnX03+mDmQ5Caeus/btb368Jcb5jYZguQ2py9AYixAw9QmA2Vgpi9AaixAbChAAou5qBKEwvbHq0EYbH+8GgS4/KMGcYft/KQj0ovdf+cTbOMT85TboU12Mnmo/Zn2TiZ2804m5g3b+Dkm+8j92efC2NfNlTD//8g+cjv7/GHbPsdkH7o/+5wbbd/QTZ0ubv9Nm4zoZjLyQQvXZNy4l00h7pXmb3bVoe5bLdKm8figdemRdUKmoBPek2flwK0D2uQdx0OSdyOrCE1CRdBRogurOAFOo6eAhI84jz7efx7FCCZ/ajhRcqd9WmeiE+Vf7z99/PDxzwfrd9nwKrOORa0s/SpySzb7wfojLI7Cs2pmy/kyTKxVUAhLlkvLXVMu2BRWmVpZLp6lLOXXMoisVSRB5YjSzKwUdoqYjXTIlpO+Y8Ku6SlbzkDnhkE94vTidLP2WLR7uSo7G/bTBZwDDw4zdZqWu8BzwzocD3hsWIfjA08Na3BcB3hoWIcD/K0KDlvm5xdHEht9SwI1FmPR+kWPmf0sTdHhBxooxdwlLpf82f8L0G1Mvw==

[gravityStar]

I've been looking at this reactor but didn't like all that steam storage. So I removed (a part of) that. Left-over energy is instead stored in the reactors, heat pipes and heat exchangers. In removing the steam storage I was able to get rid of 2 roboports and reduce the height of the reactor.



Capacity: 1108 MW Continuous.

reactors offline for (gueesstimated) 2 seconds every 200 seconds
(1120MW/202)*200=1108.91MW per second
-(2 700W to 80 000W) Lights
-50 000W Roboports
-275 000W Pumps, Lights, etc (at night)
=1 108 505 891W capacity (at night)
or 18 475 098 per tick

8 Reactors, 136 Heat Exchangers, 194 Turbines, 450K Steam capacity.

The steam tanks are only an operational buffer. Left-over energy is not stored as steam but as heat in the reactors, heat-pipes and heat exchangers instead.

Turbines dimensioned to 1 125.2 MW. Heat exchangers over-dimensioned from 112 units minumum to 136 units so that heat->steam capacity is larger then Nuclear Fuel->heat capacity. This over-capacity is necessary to efficiently use the stored heat.

To account for the fluid (steam and heat) north-south, west-east bias there are three balancer bars. One active interconnect in the center, and passive balancer bars below each row of steam tanks.

Direct lighting around reactors removed so healthy green glow is clearly visible at night.

https://pastebin.com/mNXrxKQy

[bushbaby1234]

do we have a full fixed blueprint book? or an alternative?
this really was/is my favorite reactor setup.

[impetus maximus]

do we have a full fixed blueprint book? or an alternative?
this really was/is my favorite reactor setup.

what is wrong with the one you have?

[vinnief]

I love this power plant but there are a couple of annoyances with it:

1. The pumps block the player from walking past

2. It doesn't tile quite right side by side. The large power pole and the roboport on one side need to be moved one tile

Other than that, it is pretty awesome.

[bushbaby1234]

do we have a full fixed blueprint book? or an alternative?
this really was/is my favorite reactor setup.

what is wrong with the one you have?

i thought that the change to inserters a while back messed it up, i think the specific change was that signals would completely stop them instead of letting them complete their operation, its been a while since i checked in and that was an issue at the time

[impetus maximus]

they still finish dropping what they are holding.

drop_it.png (42.92 KiB) Viewed 9914 times

[bushbaby1234]

they still finish dropping what they are holding.

drop_it.png

oh thanks!

[InflamedSebi]

I've been looking at this reactor but didn't like all that steam storage. So I removed (a part of) that. Left-over energy is instead stored in the reactors, heat pipes and heat exchangers. In removing the steam storage I was able to get rid of 2 roboports and reduce the height of the reactor.

reactors offline for (gueesstimated) 2 seconds every 200 seconds
(1120MW/202)*200=1108.91MW per second
-(2 700W to 80 000W) Lights
-50 000W Roboports
-275 000W Pumps, Lights, etc (at night)
=1 108 505 891W capacity (at night)
or 18 475 098 per tick

8 Reactors, 136 Heat Exchangers, 194 Turbines, 450K Steam capacity.

https://pastebin.com/mNXrxKQy

I was wondering why noone uses more heat pipes instead of multiple rows of steam tanks.
9 heat pipes can store around twice the amount of energy then 1 tank (9 tiles). I get that it is essential to have a steam buffer for the circuit network, since reading temperature of heat pipes is not possible.

In theory having only 1 row of tanks (18 pcs) instead of 6 (108pcs) would save 90 tanks (like gravityStar did). Those 90 tanks take up a space of 1254 tiles (19 tiles in length; 66 tiles in width; including the path and pumps).
Compensating the energy capacity loss: 90*2.425GJ = 218,25 GJ would require 436,5 Heat pipes occupying 437 tiles and save 65% of space compared to tanks. But that's not all. since additional heat pipes would need to be build at the upper part of the reactor, the exchangers would expand more to the sides. Changing shape to a more squarish/compact one.
IMO there a some additional advantages besides space usage like: heat pipes do not block walking / less pumps = less power requirement for fallback / more squareish design = easier to find a place?
the only cons I can think of, is the amount of resources needed for the pipes and the need to preheated them to 500 deg in the first startup.

Lets do the math again, to see if we even need that much energy capacity:
8xReactor = 1120MW for 200s = 224 GJ of energy to store somewhere
- 8x5GJ can be stored in the reactors themselves
=184GJ left
- 114x0.5GJ* inside heat exchangers
*(112 would be enough, but since we don't want a bottleneck in the heat->steam conversion we add +1 in general and +1 because of OCD)
=127GJ left
- 18x2.328GJ* for the single row of steam tanks
*each tank can hold 2.425GJ, BUT the circuit network needs a base amount of steam in them: for 1k steam as minimum it would be 4% of the full capacity reducing the capacity to 2.328GJ
=85.096GJ left
This leaves us with 171 heat pipes (172 for OCD) to place in the design, which are already built in without any changes, so we can already store more than 1 full cycle even without those 90 tanks.

this begs for version 2.5
8 Reactors, 114 Heat Exchangers, 194 Turbines, 450K Steam capacity. 1108.91MW continously // 1129,08MW boost

@gravityStar based of your picture you should be able to store around 1.9 fully cycles. Imma totally steal ur blueprint and shrink capacity to 1.2 cycle or something for even less space (if i can find a way to do so without breaking the circuit network ...)

[mrvn]

The problem I had (which was when reactors where freshly added and I know they fixed something with heat pipes since then) was that with more heat pipes the reactor would reach 1000° while the last heat exchanger would not have reached 500°.

I can think of two other reasons for using tanks instead of heat pipes:

1) Production costs. Isn't a tank way cheaper than X heat pipes?
2) Steam can be used directly to generate power while heat pipes need to propagate the heat to heat exchangers, need to generate steam, need to move the steam to steam turbines and only then generate power. Might take too long to switch back on when aliens attack and one needs to go from all solar to 500MW nuclear power in an instant.

[Serenity]

I'm trying to scale this up to 16 reactors, but I'm having some issues. I'm only getting 2,1GW out of this instead of 2,4GW (240 heat exchangers, 412/414 turbines). And you can see that roughly the upper right turbines on either side aren't running properly

I left out most of the pumps in the middle of the steam lines, but that doesn't seem to matter. A lot of other designs don't have many pumps either. Or are those absolutely necessary? I guess I could try some tests with hat. The outer heat pipes are colder, but still have 500°C and I'm getting 500°C steam. The outer steam pumps show less steam than the inners, but that's also the case in the 8 reactor design

EDIT:
I added a pump at after every third heat exchanger, but that doesn't help either
And making the steam turbine column less tall, but adding more columns just results in the outer ones not getting any steam

The additional horizontal heat pipes that can be added that way help distribute the heat better, but the steam situation is about the same

If I take off some turbines, the steam builds up and then more turbines are running. So it seems I'm not producing enough steam. Or it's not being distributed evenly. Could this be a pipe throughput issue between the steam tanks and the turbines?

Heat Exchangers

Nuclear1.jpg (251.16 KiB) Viewed 9485 times

Turbines

Nuclear2.jpg (278.43 KiB) Viewed 9485 times

[MeduSalem]

[...]

Don't merge pipes (like you do before and after the tanks). The pipe mechanics don't like it... at all. I wouldn't wonder if the outer turbines don't get steam because of how the steam wants to flow the wrong direction (there surely is a priority problem in the pipe mechanics which order pipe tiles (North, West, South, East) get looked at first when trying to balance out the fluids between two neighbouring pipe tiles).

Also as a suggestion... put the tanks behind the turbines and not before so that only the overflow not consumed by the turbines ever gets stored in the tanks. I have found that tanks sometimes act weird on the pipe pressure if placed in the flow-path.


If the problem still exists after that then it's most likely that you require more pumps because the pipethroughput limit is reached. Optimally you want to have almost zero steam left in the output of every heat exchanger. If there somewhere is 100 steam stockpiled in the output of the heat exchangers (or in the pipes) even if you are consuming 100% of the power the plant can give then it is most likely that you don't have enough pumps to get the steam away fast enough.

From experience I know that finding the minimum amount of pumps necessary can be quite tricky and is an iterative process of trial and error.

[Aeternus]

Fair design, but for a true megabase 1GW of power just won't suffice. There are some ideas in your control logic however that I'll pilfer and apply to my own tileable design (landfilling required - but tileable in increments).
Problem is I've not found a good way to stress-test that plant on full load...

As for tanks in the flowpath - they basically act like pipe sections with a huge storage capacity. The exit pressure is about equal to the percentage of fluid in the tank. In my experience, best practise with storage tanks is: No more then 3 in series before putting a pump row inbetween, and draining from storage tanks should be done with a pump input connected directly to the storage tank for maximum throughput. Water and steam flow rates to and from the heat exchangers to the storage tanks are usually the bottleneck in nuclear plants due to the high steam production of those exchangers. They don't like long rows as a result. With a generous sprinkling of pumps you can still achieve pretty high flow rates though.

[Serenity]

Problem is I've not found a good way to stress-test that plant on full load...

What I really like with this design is that it isolates the pumps, control and inserters from the main power grid. With that it's easy to use the Creative Mode mod and hook up a power sink to the turbine side. But if you do that with a design that has all on one grid it sucks down all the power.

And yeah, I figure the pipes around the tank are the issue. For some reason it works fine with the 8 reactor version (tried it with creative mod and it runs constantly at full power), but not beyond

[mrvn]

[...]

Don't merge pipes (like you do before and after the tanks). The pipe mechanics don't like it... at all. I wouldn't wonder if the outer turbines don't get steam because of how the steam wants to flow the wrong direction (there surely is a priority problem in the pipe mechanics which order pipe tiles (North, West, South, East) get looked at first when trying to balance out the fluids between two neighbouring pipe tiles).

Also as a suggestion... put the tanks behind the turbines and not before so that only the overflow not consumed by the turbines ever gets stored in the tanks. I have found that tanks sometimes act weird on the pipe pressure if placed in the flow-path.


If the problem still exists after that then it's most likely that you require more pumps because the pipethroughput limit is reached. Optimally you want to have almost zero steam left in the output of every heat exchanger. If there somewhere is 100 steam stockpiled in the output of the heat exchangers (or in the pipes) even if you are consuming 100% of the power the plant can give then it is most likely that you don't have enough pumps to get the steam away fast enough.

From experience I know that finding the minimum amount of pumps necessary can be quite tricky and is an iterative process of trial and error.

With the tanks after the turbines you can't use pumps though. That means they won't fill at full speed when nearly full or empty full speed when nearly empty. Means you have to power up the reactor a bit earlier.