I ran a similar off-grid system for 10 years (starting in 2003), but mine was of a somewhat larger scale. The server consumed about 400W and I had 1280W of solar panels. I started with lead-acid deep-cycle marine batteries, but later switched to sealed AGM because they did not vent hydrogen while charging. Admittedly, it was not 100% off-grid. I had a 50A 12V DC power supply on-grid that could power and/or charge the system if needed.
I switched the marine batteries out after about four years of operation.
I took down the solar panels in 2013 because the roof needed repair.
If I were doing it today, I would probably give Lithium-Iron a try. They weigh less, have higher power density, and last longer.
It would be so much cheaper to do it today, especially if you're comfortable with DIY and don't mind buying some used panels. Easier too, there are so many turnkey and DIY friendly solutions on the market now.
It's a long story. That solar plant was basically a UPS because the neighborhood where I bought a house had terrible power, and yes, they were just for the server. The root cause of the power problem was a 50-year old 4kV "last mile" which was finally upgraded to 16kV a few years before I took down the system. Power there is great right now, but maybe not for much longer because California keeps shutting down power plants. The NG plant less than a mile away will be gone next year. San Onofre is gone now too (https://www.sandiegouniontribune.com/business/story/2022-08-...). I hope to sell the place in CA soon.
I now have another house elsewhere with 4kW of solar on the roof. I would add more, but for two reasons:
1) There might be enough space up on the roof for another 2-3kW, but it's close.
2) The local laws governing grid-tie are such that I can't just add more panels. I can put up another separate system, but with a separate meter. I'm concerned that two separate systems on the same grid point trying to sync to the grid may have problems where they interfere with each other. I have read about such problems.
Regarding point #2) Absolutely yes that can be (and is) a problem for exactly the reason you say in terms of synchronising 2 (or more) PV setups to the local grid and (I would imagine) have the facility for emitting volumous clouds of magic smoke.
Fortunately the solution would appear to be use of a GPS central time clock that is fed to the 2x PV arrays individualy... if you run your own local time server you can 'skew' the clock signals to each array to get them in sync with each other and then in to a dummy load of spare lead-acids or a capacitor bank (if you want to get fancy and don't mind the 'large gray box' emitting smoke and launching itself off the wall from time to time).
Then you sync your 'mini-grid' to their slightly larger mini-grid.
Oh and when synching - never forget the old-school 4-light-bulb trick as a visual confirmation that things are synched correctly before tying in to the grid. Oh and fuses, large ones. With spares.
If you can find a friendly H.V. engineer or Linesman they might be able to tell you what clock source 'their grid' uses (GPS, Network time etc etc) which makes synching to the local-ish grid a whole lot easier if you have the exact same setup. I think the main thing is to make sure you have the dummy-load ready to feed/smooth what can be at times be a large inrush/outrush till the load equalizes.
Obviously, always seek out advice from someone that knows what they are talking about and has knolwedge of local laws, quirks of the grid etc.
Thanks for the reply. I got a few years of experience doing power systems in 1980, which came in handy when I did the CA PV system in 2003. (My embedded/web development skills also came in handy when designing/deploying the solar M&C subsystem, because no such products existed at the time.)
I'm somewhat skeptical that GPS synchronization is the real answer here. The problem is synchronizing the phase of the PV system to the grid so they can work together to power the load, allowing excess PV power to be sold back into the grid when not needed locally. To synchronize, one needs to control the phase of the PV system so it always slightly leads the phase of the grid. Determining the phase of the grid is the challenge. With two or more locally grid-tied PV systems, you cannot know the phase of the grid with much certainty because each PV's take on the grid phase is the aggregate of the actual grid phase and the phase of the other PV systems. Positive feedback and phase oscillation are likely.
Using GPS to synchronize with the grid assumes that the grid's phase is stable, but my observations over the past few years have revealed that today's grid has much less short-term stability than the grid of old. I'm assuming that the lower phase stability today is caused by the recent massive adoption of grid-tied PV arrays. So instead of using GPS time to guess the current grid phase, you really need to know what it is (locally). Aside from placing grid phase monitors on nearby houses, I don't know how to do this.
Sailors are big into LFPs these days. I just upgraded my Beneteau to a 11kWh LFP bank and it's been game changing. I cook with a 1800W inductive cooktop, use the microwave, run the desalinator, for multiple consecutive days at anchor without running a generator or the motor. Unthinkable in the days of AGM deep cycles, they'd be dead in months.
In theory they should be good for several thousand cycles, so -- a long time. As long as you don't overcharge them, drain them too low, or try to charge them in below freezing temperatures.
Also, they say that usually LFPs will last a bit longer if they're installed such that they're under a bit of sideways compression. I don't know why that is.
That only goes for prismatic cells (boxes rather than cylinders). The reason is that they tend to swell up while charging if you don't compress them so manufacturers will specify a clamping force required for the cells to keep their shape. If they change shape in such a way that you can observe it then your cells are toast so better stick to that mfg indicated compression. There are a number of articles out there that show how to do this the right way and also some that get it hilariously wrong. The direction to compress the cells in is in such a way that the electrodes remain under compression perpendicular to their surface. So compressing from the top and compressing from the front will do absolutely nothing.
You're welcome. I've been working for more than a year on a blog post about the various safety aspects of building battery packs but it's a never ending rabbithole of stuff to research and learn about so I haven't gotten around to publishing it.
Hah. Post it on HN, and let Cunningham's law help you :)
Snark aside, I feel that on HN, there are enough actual proper experts who will point out avenues of research and gotchas that a single person chasing down rabbitholes on their own is unlikely to discover.
you might be surprised what this particular single person is likely to discover on their own
one of jacquesm's previous personal rabbitholes was designing and building a wind turbine from scratch, down to cutting out the stator and rotor windings for the generator from sheet metal
No, I don't care about SEO at all. I just would like this info to be out there for hobbyists and tinkerers as well as people that are contemplating designing something with Lithium Ion based batteries in it. But because it is safety related I really want to get it all right and with little room for error I get a lot more careful about what I publish. It will happen.
My household of 3 is off-grid, running off a 15kW PV array and 24kWh of LiFePO4 batteries, with 100% electric appliances and hot water. (Location is mid-latitude Australian, plenty of sunlight most of the time)
Amazingly, no. I thought we were going to have to, but our natural usage peaks at about 20 to 30kWh per day, and we can generate far more than that on a sunny day.
Thanks to the oversized PV array, we generate power even in cloudy conditions. The PV array is also split in half and oriented east+west so it catches more sun through the entire day instead of a midday peak.
For things like clothes washing, drying, and hot water heating, those are the heavy users but can be moved to times of good sunlight.
We only run out of power if we have 2 or more days of heavy rain in a row, which only happens a few times a year.
In the Netherlands it's near impossible unfortunately to achieve 100% off-grid due to the winter months, when you basically have no solar. Maybe there are Dutchies who have somehow achieved it (with overprovisioned solar).
Do you still have a grid connection for backup or are you 100% disconnected from the grid?
Well done for going off-grid. But what do you do if you've got a whole week of rain?
A lot of people here in South Africa are keen on off grid but it's unreasonably expensive since you'll have to cater for those rainy days, and that means you'll need to build a very large system.
BTW kWh/day consumption here in the winter (which is probably typical) is around 25kWh per day.
We have a diesel generator but its only needed a few days a year; even with regular testing it only goes through about one jerry can a year, far less than we use for transport (would love an EV!)
To be frank, I've looked in my Energy supplier app (Vattenfall) and it states for December an average usage of 122 KWh (all households) and energy efficient households (solar) of 91 KWh.
I'm not OP, but I would look at this tool I found a while back. Ridiculously useful and underrated. I think someone on HN linked it, as I've not found a tool so useful via googling despite many attempts.
What you're looking for is the "Off Grid" tab, plug in your details. e.g. solar array size, size of battery, etc. Then click visualize results. After visualizing, there are a few tabs to help you make some decisions.
For your case (Netherlands, it looks like a 5.4kWp solar array and 4kWh (I assumed 5) used per day, and with a single 5kWh battery, you should be able to go off-grid :) In fact, during summery months, you could even push to the grid.
This tool even looks at cloud data based on various databases. I don't know of it's accuracy, but I'm inclined to trust it.
That's really low. Not sure how you achieve that. I think even in the Netherlands you could then go off grid with a large PV setup and diesel/propane generator for those few backup days.
I have two rooms with air conditioning but use them like 3 weeks a year (Jan/February is hot in Cape Town but you can do fine without AC). Also using gas for cooking, and heating is either via gas or wood burning stove. So your power consumption astounds me. I need to maybe just go and measure what my bits and bobs are doing. I'm afraid just my network/home server rack is >58W htough.
I have no idea why your number is so high. If it's not heating or AC, if you do run a small server room, that can explain things.
To be frank, my consumption was already quite lower, but I've made quite a lot of changes to reduce energy consumption even more.
I ran a custom x96 router/fw that I replaced with a $60 Mikrotik router that also replaces a separate switch. I've replaced more power-hungry computers with energy-efficient ones.
My existing server setup ~500Watt continously would likely take up 10KWh a day, but this is my magic trick: I don't need them 24/7. So they are off, and only turned on when required (remotely with WoL / remote power switch). The rest of what I need is on several Raspberry Pi 4B+ devices (I was lucky to buy a ton when we still could).
Our washing machine, etc seems to run non stop so it's not hard to chow a few kWh's a day. But yeah I think it's in everyone's responsibility to try to optimize and analyze one's power usage. It's also fascinating.
I like your "lazily loaded" server setup.
I have an LFP battery from EcoFlow, and now I am more worried about the inverter or BMS failing before the battery cells. End of useful life really isn't until 50% battery capacity, which could 30 years out.
I bought a low speed vehicle last summer that the previous owner had outfitted with a home made 24-cell LiFePO4 pack, complete with a cheap Chinese BMS and balancer unit. Within a week of purchase, his haphazardly installed balancer wires had melted in various places where they pressed against the high current conductors. Long story short, balancing didn’t work properly on account of the sensor wires being shorted together and, resultingly, many cells went to zero and I suppose I’m glad it didn’t also catch fire.
These things are fun, but they pack a whole lot of energy!
This is a pretty common error and the reason why in my long distance e-bike pack I spent a lot of time getting the balance wire routing done without any wires crossing.
Even Bosch got this wrong on their pannier mounted battery packs, I've opened up a whole pile of failed Bosch packs to see how they fail as a guide to my own pack design and the most common failure was water ingestion due to bad seals followed by burned out balancing wires that some how got worn out or pinched. This stuff is very critical and I would (1) never buy a pack that some other hobbyist made and (2) never sell the pack that I made to someone else.
Building batteries like these based on LiFePO4 is very easy to do, but you have to think about safety, especially if you put a battery in a moving vehicle.
If something goes wrong with these kind of batteries, it's never the batteries themselves (specifically LiFePO4) but it's always a bad connection or an undersized wire causing high resistance and thus heat --> fire.
In my case, my wires are quite oversized and I bought a proper lug nut crimping tool (and not use a vice or soldering (Don't do it)) to make the connections.
I've also really made all connections tight, double-checking them.
Yeah for sure. I am redoing all the wiring and replacing about half the cells because they went to zero… Will be over-specifying everything and making the routing super tidy. If you have the patience for that, DIY batteries are fun!
> There was a handful of new data that had been created in the intervening ~30 minutes. We pulled that diff data manually, and uploaded it to the new database.
But doesn’t that mean you had some partial downtimes? How did you manage to keep everything running during these 30 minutes where data was copied from the old DB to the new one? Let’s say there is a cron job that must be executed and its result would be stored back to the DB. Did you “pause” everything and reschedule it in the new instance once the diff copy was over?
I've been enjoying each of your posts you share here regarding this set up. I have two questions if you please?
1. Have you considered adding a small wind turbine (DIY of course) to supplement?
2. Although the costs are certainly coming down and within the realm of possibility for many, they are still too prohibitive for those that would benefit the most. Do you see any way that you could recreate your project on a shoe-string budget?
Yes, I've looked into running a small wind turbine but from what I can tell it's not worth the money or effort.
There is a reason why they build wind turbines so large and high. At small sizes, with our wind speeds, they barely produce any energy.
I can't also just put one on my balcony as I'm not allowed to (building regulations).
If I ever want to build my dream renewable energy build, I really have to relocate :)
Regarding a shoe-string budget: it all depends on your goals.
A key thing is doing everything second-hand, maybe except for the LiFePO4 batteries, unless you can find those. In the Netherlands, There's a lot of solar gear moving through our local Ebay (Marktplaats, owned by Ebay) so that's something to look out for.
Not OP but I rent and have to use my deck for solar. I only have 2 x 100W panels that sit on the ground, and that feeds a 135Ah AGM battery in a box. The box has a 500W inverter on top connected to the battery and powers a mini fridge, deck lighting, and a few other things out there. I would love to put them on the roof and my landlord would probably be cool with it, but I take it camping all the time so it's nice being able to pack them up quick.
Panels and batteries are relatively cheap (and getting cheaper) to the point that it can makes sense to have purpose-built setups.
From an actual practicality sense, the whole setup doesn't make that much sense. You could drop the files on a static site host and it would use 0 watts while not in use and a negligible amount while in use.
Another interesting case would be to use an API to get your areas power generation stats and when the wholesale pricing is low due to solar/wind generation peaks, charge your battery off the grid.
personally my UPS is connected to a 2kWh battery (with an inverter) via AC. Even when something happens to the battery to cut power too the UPs, the UPS has enough battery to last until it is resolved.
I switched the marine batteries out after about four years of operation.
I took down the solar panels in 2013 because the roof needed repair.
If I were doing it today, I would probably give Lithium-Iron a try. They weigh less, have higher power density, and last longer.