|
|
|
|
|
by namibj
3 days ago
|
|
Actually large data centers at least if done in a vaguely alirack style architecture, can do this with a decent fraction of their nominal power for very little hardware cost, as reactive power and real power add up via Pythagoras (`apparent=sqrt(real^2 + reactive^2)`) to the apparent power (rms voltage times rms current, which is what the 60Hz electronics and 60Hz transformers care about).
The first 10-ish % are nearly free. And alirack style datacenters have large 3-phase converters between the grid and some 240 (nowadays often 350) V DC bus, with the battery banks directly (with just fuses and sometimes a little bit of balancing/nudging power (think 10% of battery power rating)) on the bus, and then the servers also directly consuming from that bus. The large converters on the battery bus thus allow synthetically smoothing load transients to the grid using the batteries to smooth that power draw.
This has just minor additional wear on the batteries and a small power efficiency impact from hitting through the batteries, both of which are easily paid by anything market-rate of providing that grid service.
Because they already need the power electronics and batteries anyways, unlike a utility battery farm that at best can argue day/night load shifting of solar production as the reason for the electronics and batteries to exist. In that same spirit it's also effective to put batteries on the DC bus (between MPPT and inverters) of large solar farms, because they need the electronics anyways and it's actually reducing the required inverter&transformer capacity of the solar farm by peak-shaving. |
|