Advice on 3MW Bitcoin Mine Design

Status
Not open for further replies.
C

carsongwalker

Member
Location
Dallas, TX
Occupation
EE
Hello,

I am looking for some advice on the general direction I should go with my Bitcoin mine. I will not be the engineer or electrician responsible for the nitty gritty details, but I would like some advice on the overall design so I can work alongside those responsible for the details and begin sourcing equipment.

My background is in electrical engineering, but I am far from a power engineer, so please forgive me if I may ask some nonsense questions. I will try to be as descript as possible, but let me know if I leave anything crucial out.

  • I will be taking a 13.2kV nominal service drop, that should be closer to 13.8kV as I'm very close to the substation.
    • I intend to have a 3MVA 13,200V or 13,800V primary transformer.
  • I need to supply an initial 4 distribution panels, and would like to easily be able to expand to a total of 8 panels (ideally).
  • Each panel is rated for a total of 360kW, and requires a 415Y/240V 3-phase connection.
    • This should be up to ~500A per panel at 415V
      • ~430A at 480V
      • Each panel comes installed with a 600A fused disconnect.
    • Total of all 8 panels at full rating ~4,000A at 415V
      • ~3,450A at 480V
In order to pull this off in my required 12 week lead time I see the following design as the best option...

Design 1
  • 3MVA 13.8kV/415Y primary transformer (in-stock)
  • 2x 2500A PRL-C switchboards (in-stock)
    • Connect 2500A PRL-C inline with transformer and 4 panels. Double this for 8 panel expansion.


However, it may be cheaper to go the 480V route so that I can get away with a 4,000A main switchgear...

Design 2
  • 3MVA 13.8kV/480D primary transformer (need to find one, likely much cheaper than Design 1)
  • 4,000A main switchgear, with 4x 1600A feeder breakers (in-stock)
  • 4x 800kW 480D/415Y transformers, 1 per 2 panels (in-stock, cheap)
  • Connect 800kW 480D/415Y transformer to each 1600A feeder
    • Connect 415Y secondary to 2 panels per transformer
    • 1x 480D/415Y transformer per 2 panels
  • With a total draw from 8 panels being ~3,450A, would a 4,000A main work? (86.4%)
I'm really not sure if this is the place to be asking this, but looking for any input I can get.

One of my main questions is: Do I even need a fused switchgear if my panels have fused disconnects? Can these panels just be hooked directly to a 415Y secondary?

Appreciate you all and any insight you can give me.
 
carsongwalker said:
Design 2
  • 3MVA 13.8kV/480D primary transformer (need to find one, likely much cheaper than Design 1)
  • 4,000A main switchgear, with 4x 1600A feeder breakers (in-stock)
  • 4x 800kW 480D/415Y transformers, 1 per 2 panels (in-stock, cheap)
  • Connect 800kW 480D/415Y transformer to each 1600A feeder
    • Connect 415Y secondary to 2 panels per transformer
    • 1x 480D/415Y transformer per 2 panels
Click to expand...

An autotransfomer to go from 480V to 415V would be more efficient, but I don't know what the availability or lead time would be.
It would need to be sourced from a 13.8kV/480Y transformer.
 
What's the plan for other voltages? You'll need at least some 208/120 for smaller loads and maybe 277 for lighting. Also, the HVAC may need 480/277y depending on it's wiring and controls (e.g. for VFDs in the machines).

If you're doing lots of delta sources, how are things being grounded? (Is the 13.8k->480 is really 480/277y on the secondary with a grounded neutral? At least that's going to be a common transformer.)
 
A few thoughts.

You will be lucky to get any MV xfmr or switch gear at all in six months right now.

I would not be using any fuses at all.

If the computers require 415 V I would get that directly. Will probably need 480 V and 208/120 V too.

I would bet transformers of the size you are considering will end up with some high available short circuit currents. Will likely need maint mode switches on the feeders.

Do you need backup power? Even just for cooling?

I would bet the computers would need to already be in a warehouse state side somewhere to meet the 12 week deadline. What voltage are they? That would seem to be the place to start.
 
I just have to ask:

How much does such an electrical installation cost?
How much would the computers and ancillary equipment cost?
How much Bitcoin is floating around for the grabbing?
How much will the recurring power usage cost?
How long should it take to recoup the initial expenses?
How much does it take to recoup the continuing expenses?
 
I am also skeptical of the time frame. I'd like to see just what the lead time for the power company is.

I would avoid having two transformer transformations like the plague. Go right from primary to 415y/240.

I would try to skip the switchboards altogether, and come off the secondary of the transformer (using the transformer secondary conductor rules in 240.21) to your 600 amp main breaker panel boards.
 
I think it’s kind of a moot point. You will not get any of that within 12 weeks right now. I don’t know of anyone quoting less that 20 weeks on gear, new transformers are upward of 50 weeks right now.
 
Jraef said:
I think it’s kind of a moot point. You will not get any of that within 12 weeks right now. I don’t know of anyone quoting less that 20 weeks on gear, new transformers are upward of 50 weeks right now.
Click to expand...
One thing the OP needs to understand is the available fault current is going to be up there... Not that it's not doable, but this is going to push you into less common electrical equipment territory and unlikely to be stocked.
Might be able to squeeze below 65 k at the transformer secondary taking primary impedance into consideration. In that case, you can go from 600 amp frame down to panel board Mount branches (for example for siemens, HGB's) with series ratings, but still highly unlikely to find that stocked. I have siemens stuff on order, I'm pretty sure they are quoting 120 working days.
 
… I'm pretty sure they are quoting 120 working days.
Click to expand...
Quoting and delivering are two different things lately…

I just wasted a big chunk of time on a Bitcoin project wherein we even tried getting around delivery issues on the switchgear by using MCCs stuffed with nothing but feeder breakers just because we could get those faster, at least in the quote. Then the Bitcoin price dropped and the buyer hesitated, we lost the delivery window and now it would be 24 weeks to get the MCCs, but the switchboards went down to only 20 so we went back to that. He is still hesitating though because he wants it faster. That isn’t going to happen.
 
carsongwalker said:
One of my main questions is: Do I even need a fused switchgear if my panels have fused disconnects? Can these panels just be hooked directly to a 415Y secondary?
Click to expand...

My understanding (limited at these power levels) is that there is no specific requirement for fused switchgear, however you have several requirements which combined might lead to needing such.

1) You must provide transformer secondary protection.
2) You must provide transformer secondary conductor protection.
3) You must protect your distribution panels.
4) The protection must be able to deal with the available fault current.

The last point might dictate everything else. 3 MVA at 415V is 4200A. With say 5.5% impedance you are looking at a fault current of 76kA -> expensive OCPD.

-Jon
 
winnie said:
My understanding (limited at these power levels) is that there is no specific requirement for fused switchgear, however you have several requirements which combined might lead to needing such.

1) You must provide transformer secondary protection.
2) You must provide transformer secondary conductor protection.
3) You must protect your distribution panels.
4) The protection must be able to deal with the available fault current.

The last point might dictate everything else. 3 MVA at 415V is 4200A. With say 5.5% impedance you are looking at a fault current of 76kA -> expensive OCPD.

-Jon
Click to expand...
Yes this is what I had thrown out there Post number 9. Ok for a more detailed analysis, let's go thru it:

Your #1, the OP would be restricted to six downstream devices and they would need to be grouped in one location, not sure if that is a problem or not. The sum of their ratings cannot exceed that of a single device, so I think that part would not be a problem, even if he went up to 6 800 amp branch panel boards.

For #2, it would depend on the relative arrangement of all this equipment, that is whether the unlimited outdoor rule could be used or the 10-ft rule. The outdoor rule would be best if the panel boards could be located "nearest the point of entry". If the 10-ft rule was used, that may require upsizing the conductors a bit, depending on the primary fuse size. I think it would work with normal size conductors with a 2X primary fuse and the 6 - 800 amp panel boards, however I am not real familiar with primary fuse sizing for medium voltage transformers.

#3. Yes the panel boards would need main breakers.

#4. I took a look at Siemens just for the heck of it. If taking primary system impedance and secondary conductor impedance into consideration you can get down to 65k at the panel boards that would be a good thing. If not still definitely doable but probably going to be a bit of a price jump.

Basically it comes down to weighing the cost of a switchboard versus dealing with the restrictions on the quantity and location of the panel boards
 
This is kind of a DIY post. I would suggest you find an electrical engineer to do a one line drawing, do calcs on fault current, etc. In WA any project over 600 amps requires a one line dwg.
 
LarryFine said:
I just have to ask:


How much Bitcoin is floating around for the grabbing?
Click to expand...

From the little research I've done, I believe that the mining operation used to verify and authenticate bitcoin generates new bitcoins that the mining operation would benefit from.

I cannot answer your other questions, I do believe that this is extremely wasteful
 
Ken_S said:
From the little research I've done, I believe that the mining operation used to verify and authenticate bitcoin generates new bitcoins that the mining operation would benefit from.

I cannot answer your other questions, I do believe that this is extremely wasteful
Click to expand...
But the system has a finite number of bitcoins, not an infinite number, and each successfully mined coin increases the amount of computing that is required to mine the next coin.
 
Keep in mind the need for cooling. If you size for the computers only you will miss what is likely a significant load.
 
Status
Not open for further replies.
Top