You can build a hole that will hold up to this sort of stuff for a while, but that tends to be expensive - underground concrete work is neither cheap nor easy, and given the number of people injured in trench collapses who do it for a living, a serious underground hole isn't a good DIY project either.
That just leaves the maintenance, which also sucks for a hole in the ground. Anything you have to do on the batteries is easier with them on or above the ground, instead of underground in a hole. For lithium, it's fairly low maintenance but you may have to replace some BMS boards every now and then, manually check balance, etc, and for lead it's water every few months, unless you go with sealed, which are "maintenance free," but also "maintenance and status check impossible," so you really just set them and run them to failure - you can't monitor condition like you can with flooded.
It's the sort of solution that, technically, would work - it just ends up being radically more expensive and a hassle than it's remotely worth. Think "Spending $5000 to save $20/yr" sort of thing.
You see the same trend with things like the complex solar trackers that were common in the 80s and early 90s. With the cost of panels, optimizing every bit of power from them was critical, so spending an awful lot of money on support infrastructure to optimize $10/W panel production made quite a bit of practical and financial sense.
Today, with panels at $0.50/W, very few people bother with trackers anymore (for off grid installs). Most things you could improve with a tracker, you can improve with less effort and less money by just hanging more panels. Bonus: With more panels, you get more production on cloudy days when panel orientation doesn't matter (because the sky is more or less evenly lit).
I've got south facing, east facing, and west facing panels on my office, of two different brands, all fed into a single MPPT charge controller input (I do have some blocking diodes to prevent backfeeding a shaded string, though). It's not nearly as "efficient" as some other options, but it's a whole lot cheaper, and gives me more power on the low cloud days that are the rough ones. The rest of the year, I have so much power out there that I'm demand limited, not supply limited (actually, inverter limited - if I added some 48VDC loads, I could use another 5kWh/day easily).
Get a plastic storage box, drill holes for cables and seal them with caulk. Put inside a treated lumber box for extra crushing resistance. Run cables through metal conduits to avoid mice and moles chewing on them.
Rent a backhoe, and you are done in half an afternoon.
All of that said, I dont know that I would actually go that far- you avoid the battery freezing, but then you get to worry about it cooking. For an off-grid cabin that isn't constantly heated in winter, it isn't the worst idea.
No issues so far, two years in. It’s not terribly cool in the hottest days of summer and does require a small heater if I want to keep it above freezing through winter, but in terms of being difficult or expensive to build, not at all.
I know it’s only lighting as a load, but I’ve heard standard LED light bulbs will happily run as low as 36VDC, as they chop the voltage down to that level anyway.
It makes me wonder how many “120VAC” devices with switch mode power supplies could be plugged into 48VDC without any thought at all. And not just electronics like a laptop or non-plasma flatscreen, but everything is moving in this direction:
Like… what’s my Variable Frequency Drive “inverter” refrigerator motor voltage really running at? In theory it should happily run off 120VDC at least. Don’t do that, but somewhere under 110VDC should work fine. And it’s inherently soft start and easier on one’s low voltage wiring than an on/off duty cycle fridge.
Of course I recommend none of this officially.
Active power factor correction circuitry will have a fit, but... eh, whatever. It's DC now, nobody cares!
A 120VRMS AC signal peaks around 170V, and most (not all, but most...) switch mode power supplies are auto ranging, so they'll tolerate from about 100VRMS to about 250VRMS - which is near 400V peak. If you can get 300+VDC into them, they'll be quite happy, but the boost converters to do that are a pain to find, and by the time you build one, you may as well just get a 120VAC inverter and call it good.
The main thing to be concerned about, and the reason to jack the input voltages about as high as you can, is that almost every switch mode power supply has a full bridge rectifier on the input side. From an AC input source, each diode has a 50% duty cycle. Put DC on it, now two diodes are at 100% duty cycle, two are at 0% duty cycle. How much overhead was there in the system? Well... you'll find out! However, a system capable of running and not overloading the diodes at 50% duty cycle at ~100VAC should be totally fine in terms of diode heating at 300VDC - just, perhaps not at 100VDC.
All I could think about was the (potential) improvements on capacitor lifespan since they won’t have to smoothen, but wasn’t sure if their electrolyte could still heat/dry if it was actually the connection to power that makes them hot.
Thinking further, an internal fuse might blow with the (expectedly) larger current draw at 48V.
And yeah on input wiring, you’re pulling more current at 48V.
But as soon as you've got some stuff that requires an inverter, you may as well just light the inverter up for everything. I've got the equipment to do a separate, lithium-backed "DC rail" in my office that would be around 40V, and then buck it down for various devices, run my routers direct on it, and... I've never gotten around to doing it, because until literally everything is on that, I'd still have to run the inverter for things like system sleep (I sleep most of my computers overnight). I'm fiddling watts around and it's just not worth it unless I can actually shut the inverter down entirely. Unfortunately, boosting up to 300VDC isn't cheap or easy.
I've wondered if you could run some of that stuff straight off solar during good sun, but I've never really wanted to subject my computers to that sort of abuse.
That’s where I’m coming from: trying to convince a family member to setup a DC system so the inverter can stay in sleep mode during evening/overnight, because the efficiency is horrendous at small loads. it’s really academic in kWh savings per day, but it saves you the most when sunlight’s the least and reduces battery cycling. But if you’re running an office, the inverter is pretty much running a good load or off, I think.