Earth Notes: Off-grid Battery Bank Replacement (2021)

Updated 2023-01-20.
10 year old battery bank was at about 10% of original capacity. I went for a simpler and smaller and cheaper replacement.
A decade ago I upgraded from a 40Ah battery to a ~400Ah battery bank, which is now worn out. But I now don't think that I can fully use that much capacity even with the huge increase in PV. So I'm going for a smaller and less complex battery this time around, but still lead-acid. Lead-acid works, is recyclable, and I don't yet trust or understand lithium BMSes!

2021-05-03: New Battery?

Yet again, I'm considering replacing my gel LA (lead-acid) set now over 10 years old.

I don't think I'll go for another ~400Ah because I don't think the charging energy is there in winter to justify it. With a minimum load of the RPi(s), much less than 100Wh/d capacity is needed to carry it, maybe only 50Wh/d. A completely sunless week suggests ~500Wh required usable capacity, ie down to ~50% DoD. That could be ~100Ah at 12V.

(If I do have excess above the bare minimum then taking other loads off-grid such as ~300Wh/d of router is good. But 50Wh/d to move off-grid through the evening peak is also acceptable.)

2021: ageing 12V 4x 99Ah gel battery bank
Ageing 12V 4x 99Ah gel battery bank after 10 years' service.

One suggestion is a drop-in single Victron Energy GEL Deep Cycle Battery 12V 220Ah BAT412201104. (£429 including VAT, shipping £22 including VAT.)

Another more adventurous suggestion by "Scruff" is one (or two) Firefly 12V 116Ah G31. (£410 each including VAT and shipping.)

If you're shallow discharging then one firefly with 116Ah has more energy over ownership than a Victron 220Ah (you can really "abuse" firefly, 80% DOD is perfectly acceptable).

In this case Scruff is suggesting setting the SS MPPT to flooded mode, but with equalisation turned off. Maybe the internal setpoints could be tweaked a little bit too. In particular the SS MPPT temperature compensation is -31mV/°C, vs the Oasis suggested -24mV/°C. (Both with a reference of 25°C.) Scruff created an MSView file to fix them. DIP switch 1 has to be on, and the jumper inserted.

Given my estimate of the current old set current usable capacity of ~300Wh, a single Firefly at ~700Wh usable would be an improvement.

There's some other minor issues like needing new M10 cables unless I drill out the old ones. And disposing of the old batteries (and my original 40Ah!)...

(Shipping is included in the Firefly price, so I could order a first, get the initial cabling right, then order a second soon after and get the cross-straps in place.)

Maybe the end of summer is the time to do it, given that there is less work for the batteries to do until autumn.

2021-05-14: Tanya and Victron Gel

I made a gentle enquiry with Tanya about the Victron Energy GEL Deep Cycle Battery 12V 220Ah BAT412201104.

The responses were polite and fast and informative!

Tanya said that there would be a fixing kit, nuts/bolts with the battery, and that my existing M8-lug cables should be fine. Lead time is currently 7-14 working days from order. But any date up to autumn would be fine.

The advantages of this less exciting option over the Firefly might be:

  • The "bank" would be just one battery with no cross-wiring.
  • The solar controller settings can remain unchanged on "gel".
  • Most or all of the tuned parameters in my powermng code can probably remain unchanged.
  • No new cables to buy or route.

Given that I don't know how long we will remain in this house given the estate redevelopment, maybe this is not the moment to try to maximise longevity.

The Firefly would however give me experience of novel AGM, so maybe I should seek out the opportunity to do that elsewhere if not here.

2021-09-02: Victron Gel Arrived

Out of cowardice/simplicity, today I have a shiny new 220Ah Victron 12V gel. I shall install it over the next few days. I shall dispatch the old bank (and the 40Ah) via a local scrap dealer.

I ordered from Tanya yesterday afternoon.

No one seems to want to take my old LiPO4...

The new battery is very heavy (66kg). I'm likely to need help to get it installed to avoid injuring it or me.

The DEKA batteries being removed each weigh 28kg, ie 112kg in total.

The new battery does indeed come with nuts/bolts for the terminals.

(Terminal covers from Bimble are on their way too.)

2021-09-03: Swap Shop

The plan:

  1. [Done ] Move 40Ah (and Li) battery out ready for collection, including recovering polyfuses, connector blocks, etc, and disentangling any remaining connections. Interestingly, 40Ah gel showing 11.9V having not been connected for many many months, maybe years. Li showing ~4V.
  2. [Done ] Move RPi servers and router to mains power until done, power disconnect button from shed pressed to avoid any transients getting into the house.
  3. [Done ] Disconnect the PWM controller inputs (PV) and then outputs (to battery) until (nearly) everything else done.
  4. [Done ] Disconnect and move from the shed one or two of the DEKA batteries if possible without bringing everything else down.
  5. [Done ] Disconnect the MPPT controller inputs (PV) and then outputs (to battery).
  6. [Done ] Disconnect and remove remaining DEKA batteries including temperature sensor (was on a -ve terminal) for MPPT controller.
  7. [In part!] Empty shed of everything that might get in the way of heavy lifting.
  8. [Done ] Move new 220Ah into place. Needs help! Jean and I managed together, with a stop half-way up.
  9. [Done ] Connect up PWM and MPPT controllers and MPPT temperature sensor and fit terminal covers.
  10. Observe battery (and MODBUS and powermng) behaviour until stable.
  11. [Not yet done] Take opportunity to observe PWM inputs, ie all 12V PV panels working?
  12. [Done ] Move RPi servers and router back from mains power.

The terminal covers from Bimble arrived and are huge! A mouse could comfortably set up home in each... Not quite what I expected, and may not exactly fit the terminals as intended, but I think that they can be pressed into useful service.

Speaking of mice... Something had evidently tried to make a nest behind the battery bank, and some of the polystyrene packing I had as thermal insulation was chewed up. But no corpse or other damage that I could see.

Some stats running up to full reconnection (the sky is overcast):

2021/09/03T12:40:06Z AL -1 B1 0 B2 -1 P -1 BV 0 ST VL - L A1P -1 B1T -999 UC 0 A1V -1
2021/09/03T12:50:06Z AL -1 B1 4536 B2 -1 P -1 BV 4536 ST VL - L A1P -1 B1T -999 UC 0 A1V -1
2021/09/03T13:00:06Z AL -1 B1 0 B2 -1 P -1 BV 0 ST VL - L A1P -1 B1T -999 UC 0 A1V -1
2021/09/03T13:10:06Z AL -1 B1 4431 B2 -1 P -1 BV 4431 ST VL - L A1P -1 B1T -999 UC 0 A1V -1
2021/09/03T13:20:06Z AL 1631 B1 13093 B2 -1 P 0 BV 4650 ST OK D E A1P 21320 B1T 24 UC 100 A1V 26560
2021/09/03T13:30:06Z AL 1481 B1 13389 B2 -1 P 67 BV 0 ST OK D T A1P 19885 B1T 20 UC 100 A1V 26508

The BV value is just noise since the RPi is floating wrt the battery until I bring it back off mains.

There was one bit left over! I'd failed to connect up the PWM battery -ve. I think that it had an indirect connection as I have connected the negative sides together of the two panel sets. Here's hoping I haven't let out any magic smoke. I have now connected it, but it needs to be re-done, better, IMHO.

I have tentatively powered the Mac from the off-grid system as a test. The voltage is still a bit marginal for that, though.

I am pleased to see the new battery soaking up more than 60W via the MPPT alone, even in far from optimal conditions . Total power from the primary array passed 100W while feeding ~40W to the Mac.

First nominally back-to-normal log entries:

2021/09/03T15:30:06Z AL 2699 B1 14158 B2 -1 P 4389 BV 13995 ST H D e A1P 38155 B1T 20 UC 100 A1V 27591
2021/09/03T15:40:07Z AL 2740 B1 13965 B2 -1 P 16870 BV 13653 ST H D T A1P 38200 B1T 20 UC 100 A1V 27591
20210903 battV
Off-grid storage heart-transplant chart!
20210904 3 battV
Off-grid storage heart-transplant chart continued...
off grid storage gel LA and LiFePO4 since 2007
Off-grid storage since 2007, all nominal 12V batteries, now retired: LiFePO4 20Ah with BMS, lead-acid gel 40Ah, 4x lead-acid gel 99Ah.

2021-09-05: Tip Top

My shoulders were very sore last night, after Friday's lead lugging!

Now after yesterday's gloom we have some sun today and the new battery hit FULL soon after solar noon and moved to float charge:

2021/09/05T12:20:06Z AL 2380 B1 14194 B2 -1 P 15713 BV 13881 ST VH D V A1P 33867 B1T 19 UC 100 A1V 32071
2021/09/05T12:30:07Z AL 1839 B1 13889 B2 -1 P 15445 BV 13556 ST F D F A1P 25562 B1T 19 UC 100 A1V 32117
2021/09/05T12:40:06Z AL 1955 B1 13889 B2 -1 P 15640 BV 13547 ST F D F A1P 27072 B1T 19 UC 100 A1V 32126
20210905 battV
A full-fat battery bank!

I stand ready to tweak parameters for powermng such as voltage thresholds and impedances.

I will also do a little more "winterising" preparation, such as better thermal insulation around the battery. I have had a stab at the MPPT's remote temperature sensor on the -ve post, wrapped again with sponge to reduce thermal losses to the air.

Note that the polystyrene and polyethylene packing for the battery has been repurposed as extra thermal lagging. The cardboard will be recycled shortly.

Victron LA gel 12V 220Ah in place 20210905
Victron 220Ah in place, partly insulated from the coming cold.

Now that powermng has seen the bank FULL, it should be much more willing to dump. That is, take the router off mains even when the grid is not at peak intensity.

The off-grid system is now supporting the router, and charging my Mac and phone:

2021/09/05T12:50:06Z AL 2035 B1 13889 B2 -1 P 15917 BV 13547 ST F D F A1P 28280 B1T 19 UC 100 A1V 32236
2021/09/05T13:00:06Z AL 5402 B1 13880 B2 -1 P 66430 BV 12205 ST F D F A1P 78573 B1T 19 UC 100 A1V 31269
2021/09/05T13:10:06Z AL 4958 B1 13880 B2 -1 P 61239 BV 12091 ST F D F A1P 68774 B1T 19 UC 100 A1V 31238

(I still have the retired batteries to have properly recycled...)

My hope is that for the next few days at least while we have sunshine, but maybe most of the next month or three, that the router is kept off-grid most of the time.

So the upgrade only took about four months or four years, depending on when you think that the bank should have been replaced!

2021-09-06: Top Cat!

This morning from the chart it is clear that the router has stayed off-grid all night, ie continuous dumping.

20210906 battV
Off-grid router all night...

Also the state of (usable) charge did not drop below 100%:

2021/09/06T00:00:06Z AL 0 B1 12665 B2 -1 P 14046 BV 12424 ST OK D e A1P 0 B1T 21 UC 100 A1V 0
2021/09/06T00:10:06Z AL 0 B1 12665 B2 -1 P 14540 BV 12416 ST OK D e A1P 0 B1T 21 UC 100 A1V 0
2021/09/06T05:40:06Z AL 0 B1 12626 B2 -1 P 14217 BV 12354 ST OK D E A1P 0 B1T 20 UC 100 A1V 16657
2021/09/06T05:50:06Z AL 37 B1 12626 B2 -1 P 14647 BV 12354 ST OK D E A1P 453 B1T 20 UC 100 A1V 20798
2021/09/06T06:00:06Z AL 100 B1 12626 B2 -1 P 14584 BV 12354 ST OK D E A1P 1239 B1T 20 UC 100 A1V 21631
2021/09/06T06:10:06Z AL 141 B1 12626 B2 -1 P 15227 BV 11477 ST OK D E A1P 1722 B1T 20 UC 100 A1V 22025

(This morning is misty, so 'real' autumn, though warm and bright later.)

Early afternoon all six batteries, including the LiPO4 and 40Ah LA, were picked up for recycling.

2021-09-07: He's the Pip

20210907 2 battV
Off-grid router all night again, 388Wh generation from primary array, 365Wh load yesterday.

It's all still going to plan. The router kept off-grid for another full night. 388Wh generation from the primary array, and 365Wh load yesterday.

A pre-dawn log sample on what is forecast to be a scorching 29°C day:

2021/09/07T05:00:06Z AL 0 B1 12665 B2 -1 P 13197 BV 11521 ST OK D e A1P 0 B1T 22 UC 100 A1V 3608

Note the little bit where dumping stops, which is due to 's' status. That lets the off-grid battery catch up on charge when the house is already generating significantly and exporting to grid. The reasoning is that the 12W of potential extra exports is not significant, and maintaining battery health is more important. This is still good policy, IMHO.

But given the shiny new battery I am keen to keep dumping when reasonable. So I have made powermng a little less likely to go into 's' mode unless the system has not seen decent sunshine in a while. It may need a further tweak in that direction. This tweak would have also been fine with the old failing bank.

2021-09-12: He's the Championship

It's been nice to see the MPPT controller soaking up nearly 200W at moments. No longer having its style cramped by the previous ageing battery bank.

20210912 battV
MPPT controller now capturing peaks of nearly 200W.

2021-09-15: (He's) The most Tip Top

The new battery has already dealt with very bright days, very dull days, and dull days with bursts of bright direct sunshine, keeping about 0.25kWh/d of 'dump' load (the Internet router) away from the grid.

20210915 battV cm
Longer-span chart showing clearly the installation and the changed general before-and-after characteristics.
20210915 EOall
Chart showing energy capture by MPPT controller over several years.

2021-09-16: Top Cat

Having not managed to get up to VHIGH for the last couple of days, the dump load has just been returned to the grid for the first time in several days to give the battery chance to charge. It may be that I should extend still further the time since VHIGH for this to happen, given that it was actually quite sunny yesterday.

2021/09/16T09:00:06Z AL 865 B1 12665 B2 -1 P 14692 BV 12442 ST OK D E A1P 10917
B1T 18 UC 92 A1V 27640
2021/09/16T09:10:06Z AL 914 B1 12702 B2 -1 P 14366 BV 12451 ST OK D E A1P 11566
B1T 18 UC 98 A1V 27292
2021/09/16T09:20:06Z AL 1020 B1 12732 B2 -1 P 14490 BV 12468 ST OK - s A1P
12970 B1T 18 UC 100 A1V 27445
2021/09/16T09:30:06Z AL 1073 B1 12894 B2 -1 P 2992 BV 12767 ST OK - t A1P 13816
B1T 18 UC 100 A1V 27445

I have wilfully undermined this by putting my phone on to charge which is a similar load!

By solar noon VHIGH had been hit:

2021/09/16T11:20:06Z AL 3984 B1 13959 B2 -1 P 4663 BV 12459 ST H - t A1P 46912 B1T 18 UC 100 A1V 27234
2021/09/16T11:30:06Z AL 1445 B1 13209 B2 -1 P 3871 BV 13082 ST OK - t A1P 19085 B1T 19 UC 100 A1V 25440
2021/09/16T11:40:06Z AL 10891 B1 14209 B2 -1 P 3951 BV 14100 ST VH - t A1P 154248 B1T 18 UC 100 A1V 31250
2021/09/16T11:50:06Z AL 3520 B1 13584 B2 -1 P 4470 BV 13363 ST H D e A1P 47878 B1T 19 UC 100 A1V 24662
2021/09/16T12:00:06Z AL 9381 B1 14188 B2 -1 P 15153 BV 13934 ST VH D T A1P 133170 B1T 19 UC 100 A1V 31211

I have pushed out time-since-VHIGH to 3d, and time-since-FULL to 30d based on typical equalisation schedules. This is quite aggressively in favour of keeping the 'dump' load off-grid, providing SoC thresholds allow. These times are now safety/health boundaries.

As of mid-afternoon FULL was achieved, edge of the previous time-since-FULL threshold (as well as tripping the VHIGH one earlier). There has been quite a lot of sunshine amongst the gloom suggesting that as battery-health backstops they were set too conservatively.

2021/09/16T14:50:06Z AL 5175 B1 14158 B2 -1 P 16636 BV 12916 ST VH D A A1P 74180 B1T 20 UC 100 A1V 31153
2021/09/16T15:00:06Z AL 4177 B1 13849 B2 -1 P 18378 BV 12591 ST F D F A1P 57934 B1T 20 UC 100 A1V 29300

2021-09-25: Gloom

The 25th was a gloomy day with poor generation (~80Wh), only net-charging for a few minutes. The previous day, FULL was hit (~400Wh generation).

Yet the router was kept off grid all day, like a champ.

It was just me at home, and there was little import from the grid with the Enphase carrying most load. In effect, today transferred ~250Wh from the off-grid store to the Enphase grid-coupled store.

20210925 battV
Barely any net charge, but ~12W kept away from the grid throughout.

2021-10-01: Gloom

After continuing gloom the router dump load went back on grid in the wee hours. It was off-grid again four hours later. For some of the time in the interim, the Enphase carried it, so still somewhat off-grid! This is an expected pattern for this time of year.

2021/10/01T03:30:06Z AL 0 B1 12476 B2 -1 P 13113 BV 12108 ST OK D e A1P 0 B1T 15 UC 60 A1V 0
2021/10/01T03:40:06Z AL 0 B1 12473 B2 -1 P 13870 BV 12038 ST OK D e A1P 0 B1T 15 UC 60 A1V 0
2021/10/01T03:50:06Z AL 0 B1 12473 B2 -1 P 13172 BV 12030 ST OK - S A1P 0 B1T 15 UC 59 A1V 40
2021/10/01T04:00:06Z AL 0 B1 12589 B2 -1 P 1864 BV 12345 ST OK - t A1P 0 B1T 15 UC 65 A1V 22
2021/10/01T04:10:06Z AL 0 B1 12589 B2 -1 P 2745 BV 12363 ST OK - m A1P 0 B1T 15 UC 66 A1V 10
2021/10/01T08:10:06Z AL 469 B1 12641 B2 -1 P 2933 BV 12424 ST OK - m A1P 5919 B1T 15 UC 68 A1V 26368
2021/10/01T08:20:06Z AL 870 B1 12742 B2 -1 P 2778 BV 12556 ST OK D e A1P 11068 B1T 15 UC 81 A1V 27448
2021/10/01T08:30:06Z AL 585 B1 12574 B2 -1 P 14712 BV 12161 ST OK D T A1P 7369 B1T 15 UC 70 A1V 26465

2021-10-09: Update

Autumn is definitely with us bringing many dull days with low insolation.

20211008 EOall
Chart including energy capture by MPPT controller into October.

But today has some (hazy) sunshine and the charger maxed out at 15A. Eventually more than 600Wh was generated/absorbed.

2021/10/09T11:30:06Z AL 15000 B1 14066 B2 -1 P 15712 BV 13749 ST H D h A1P 210716 B1T 16 UC 100 A1V 28989
2021/10/09T11:40:06Z AL 15000 B1 14011 B2 -1 P 15819 BV 13819 ST H D h A1P 211682 B1T 16 UC 92 A1V 28071
20211009 battV
Catching up after a dull day, peaking at 15A/212W; ~612Wh captured.

2021-10-21: Not FULL

Even on a relatively sunny day, there does not seem likely to be a long enough day to hit FULL now.

2021-10-31: FULL

Well, today after a few days doing 'peak'/'emergency' dumping only, it hit FULL. Now not FULL since Sun 31 Oct 15:00:08 UTC 2021, previously 2021-09-24.

After further reading up, I'm extending the time between the system wanting to see FULL else severely curtailing dumping from 30 days to 90 days. I will however further restrict the depth of discharge for dumping for a long time since VHIGH or FULL at their timeouts.