AJW22's modular 3d printed 60kWh PowerShelf

ajw22

Member
Joined
Nov 16, 2018
Messages
733
My PowerShelf project has come a long way sinceI started collecting 18650 cellsabout 2 years ago. This is the latest status (April 2020):

Background:
I've had my gridtied system with feed in tariff installed a few years ago, which unfortunately is nearly useless on cloudy days and entirely useless during the night.
At my companyI had a drum full of discarded powertool batteries,and decided to put them to good useafterstumblingacross a Youtube video featuring somequirkyAussie who built himself a powerwall.
Later switched to discarded ebike packs containing mostly 2Ah Sanyo cells. While not free, it's still cheap and there's a near endless supply.
While spendingseveral months harvesting cells as well as reading up on various projects, I figured that I'd need years to complete the full system I wanted. So the logical approach for mewas to construct a modular system that I could expand byadding PV panels / chargers / batteries / invertersas I went.


=== Overview of current status /[planned max]===
* [DONE!]9.44Wsolar panels
* [DONE!]3x eSmart3 60amp solar chargers configured to max total 120Amp
* Batteries in parallel: 6x 14s108p;nom 60kWh, 45kWh ACusable /[10x 14s108p, nom 100kWh]
* 6xChinese"Smart BMS 60A" with bluetooth,RaspberryPi3 + Grafanamonitoring
* 3x 1kW grid tie inverter with limiter /[4x 1kW]
* [DONE!] 1x 1.5kW emergency standalone inverter for during blackouts


=== The details: Solar panels===
Canadian solar 265W on 4x racks @6 panels. Connected in4p3s configuration. EacheSmart3 charger is attached to 2 DIY racks. Racks are securely anchored to the concrete.

image_rcsxcy.jpg




=== The details: eSmart3 60amp solar chargers ===
The 3 devices areeachconfigured to charge ata maximum of 40Amp (combined total 120Amp)due to the limitations of the batteries, and57.12V(4.08V*14) to extend battery life. Quite happy with these units, though one cooling fan decided to disintegrate and needed to be replaced.

image_pmxhir.jpg


image_hztdqs.jpg



=== The details: Battery packs ===
Made of custom designed 3d printed parts. I compromised severely on the electrical design aspects to maximize the serviceability and ease of construction/replacement. I've made various improvements to the design, and it's now in the 9th Version. I've attached the source .SCAD file below. You need a free CAD software called OpenSCAD edit it (it's just a text file), render ("F6"), and save as .STL (File->Export->). Edit the bottom 10 lines to suit your needs, it should be pretty self-explanatory.

PRO:
* a pack can be taken out by undoing2 wood screws and disconnecting the XT60 connector. No need to fumble with theBMS lead as it'sintegrated into the XT60 connector.
* Apart from removing/reattaching the fuse wire,a cell can be replaced without tools simply by pushing them out. No prying apart, hammering, etc required.
* bus bar is made of just 1 length of partly stripped5.5mm^2 cable. No twisting, powertools or bending contraption required
* hardly any possibility of accidentalshorting, even during maintenance or whenbumping into itwith metal tools stuck all over your body.
CON:
* Cells will(dis-)charge unevenly under high load, so limited to 30A
* Cell voltage monitoring might be inaccurate under high load

image_ivbrwx.jpg


image_fizzaq.jpg


image_eexiwf.jpg



=== The details: 18650 cells ===
Initially I started off with a free supply of oldpowertool batteries. Unfortunately, they typically are 1300~1500mAh low capacity/high drain cells - the exact opposite of what a powerwall needs. After exhausting that supply, I switched to old ebike cells. Now I almost exclusively use1800~2400mAh Sanyo cells.
I check the capacity and test for self-dischargeevery cell, but do not sort them in any way. My latest 108ppacks are built with basically randomly selected cells, although I try to put in some low capacity 1300mAh cells into the front ofevery pack. The packs generally end up with remarkably similar capacities, but if they do not, I can easily swap some ofthose low capacity cells with higher capacity ones(or vice versa, if the pack capacity is too large) to equalize capacities.

=== The details: BMS/wire harness ===
I useone Chinese 14s 60ASmart BMS purchased on Aliexpresson each battery. The sensor/balancing leads are integrated into the XT60/harness so as to simplify maintenance. There is a 30A fuse inline in the harness for extra safety.
PRO:
* Cheap
* Balancing can be set to start at any voltage and diff. Currently configured start if at3.9+V andmore than 0.015V difference.
CON:
* Can balance/bypass at only50mA. This is so farproving to be enough, except whenthere are faulty cells. On the plus side, faulty cells will not go undetected.
*Initial balancing / balancing after maintenance can take days, if not weeks.

image_evwdds.jpg






=== The details: 1kW grid tie inverter with limiter===
I went with these instead of hybrid inverters, because I didn't want to rewire the whole house breaker. This just requires a dedicated line to the breaker panel and aclamp meter on the main line to prevent feeding back into the grid. At the moment I have 3 of those throttled to max 750W. While they do not feed back into the grid,they often do not share the load equally. They are currentlyconfigured to slow down / stop when the battery voltage reaches 48.3V (3.45V*14)

2019-11-26 update: The Inverters are now connected to the breaker boxvia a TP-Link HS105 Smart Wifi Plug. This allows the Raspberry Pi 3 (below) to disconnect the inverters during off-peak hours (11pm~7am, 1/3 price!)if the batteries are kinda low.I hope to implement a smarter system in the future that takes into account weather forecast data
2020-04-29 update: Changed the batteries low threshold during the winter months to 3.7v/cell. This allows the batteries reach 3.9V/cell on the occasional sunny dayand do some balancing.


image_vendds.jpg



=== The details: Raspberry Pi 3 + Grafana monitoring ===
This hasproven extremely useful. With the pack levelvoltage vs timegraph, I can easily see when a pack has developed a problem - even distinguish between a blown fuse, self-discharging cell, or reduction in capacity due to wear.
I've taken the snapshot below a day afterreplacing/ reshufflingsome cells. Not enough time for the balancer to properly top balance.

image_uhuoya.jpg
 

Attachments

  • BatteryPack9.zip
    1.9 KB · Views: 716
Attached my OpenSCAD source and .STL file to the top post. You can change the pack size by opening the source file in OpenSCAD (freeware) and just editing 1 number (the "cells" variable).
It now creates 70P brackets, but you can create brackets as short as 10P or as large as your printer allows.

I print using PLA, though ABS with its higher melting point may be a safer choice in case of catastrophic cell failure.
 
I am impressed. I am in the process of building packs of 14S80P with the single holders (about 5kWh per pack). This give me packs of 400mm by 560mm which can be located easily next to each other. So I do not need to go to much in to the height (as I life in a earth quack prone environment). The first pack is fully packed and the next couple of packs are underway. I am looking to use the smaller BMS to have every pack separate controlled like you probably do. I want to start hooking them up when I have six packs (approx. 30 kWh). My current issue is to fin what kind of inverter to use.
 
A few words on Raspberry Pi 3 + Grafana monitoring, and its uses.

I thought about using rePackr to assemble thepacks for my first 14s92p battery. Turns out I'm too lazy to catalogue1288 cells, much less sort/fetch the exactones rePackr says I need. So instead I went the sort-into-100mAh-buckets route.
For the second battery(14s108p) I couldn't be bothered to do even that. The third battery(another 14s108p) was assembled using totallyrandom cells, though I capacity & leak tested all of them.

After 6 months of operation, I finally got around to installing a remote monitoring system.
The 3x chinese "Smart BMS" each camewith a bluetooth dongle. The (extremelypoorly written)specificationsare listed on their websitehttps://www.lithiumbatterypcb.com/, and there are several projectson the internet if you search for "smart bms bluetooth programming".I couldn't get any of them to run properly, so I wrote my own python script to connect to andquery 3 BMSsand once a minuteupload the data to alocal (RaspPi3 that is)Postgres DB server and graph the data usingGrafana.
The first time I drained to 3.2~3.35V, it was immediately noticeable that the packs had mismatched capacities. One pack had broken fuses (bad soldering), somehad cells withelectrolyte leaks, others were low/high capacity purely due to badluck.

(ignore the noises/blips. Just software glitches I've nowmostly sorted out)

image_rojpdt.jpg


After fixing the problems as well as shuffling some low/highcapacity cells to balance out the capacities:

image_ydygej.jpg


So not only does this system help me identify and fix issues, it also allows me to build my packs in the most lazy way and sort out anycapacity balanceissues later :)
 
In this time I am to building in 80 cells by pack with plastic pieces 4x5 for 14s80p
If I could see before your very good conception I have to choose your principal concept
Congratulations very good choices
Just a question you have sold fuse wire in series with cells in parallel ?
Bye
Eric
 
Walde said:
What kind of 3D printer do you have?
Did you use PLA or ABS?

I used PLA ona CR-10 printer with the small 30cm x 30cm bed to print my current 14S92P battery brackets.
My next battery will be 14S108P, the maximum my printer can handle.
Quite confident that this design could be expanded to much larger packs if you have a larger print bed.


eric.peton@laposte.net said:
In this time I am to building in 80 cells by pack with plastic pieces 4x5 for 14s80p
If I could see before your very good conception I have to choose your principal concept
Congratulations very good choices
Just a question you have sold fuse wire in series with cells in parallel ?
Bye
Eric

Yes, I have 2 cells sharing 1 fuse wire that breaks at ca 4 Amps.
My initial plan of using axial glass fuses on each cell seemed quite difficult to execute space wise, so I used AWG 40 wires as fuses. I got the wires by separating the strands of a AWG 26 cable.
I am _maybe_ regretting this decision a little.
The longish fuse wire (1cm ~ 6cm * 2 sides) has a lot more resistance than the axial glass fuses (fuse part is just 4mm), causing an estimated 2%~4% energy loss at max pack load of 25 Amps. Twice that if you include charging and discharging. Probably not a big issue once I have my 2nd parallel battery online.
The next time, I will probably use 32 AWG on at least the negative side. That should reduce losses significantly while still giving some backup protection. Perhaps also use 2 strands of 40 AWG on the positive side.
Also, 40 AWG wires are a pain to handle and solder. And they break easily when stretched, so I needed to add some slack.
 
I really like the 3d printed cell holders but would prefer a slightly different configuration... are you in the mood for doing some STL design work for others with 3D printers but not the skills to design?
 
ajw22 said:
* Finally settled ona very simple2x 2mm^2 straightsolid copper bus bar, which should haveno problem with my max 25A use.

You are aware that 50% of your cells are fused with 2x the current right?
In case of a short an outer cells will only see maximum the fuse-rating.

But an inner cells will se 2x this current before both leads blows.
 
ajw22 said:
Hanssing said:
ajw22 said:
* Finally settled ona very simple2x 2mm^2 straightsolid copper bus bar, which should haveno problem with my max 25A use.


Todo:
1. Move batteries to bolted down dedicated rack(80% done)
2. finishanother14S104P battery (70% done)
*. add another 1kW GTI
*. harvest and build more 14S104P batteries (0%)
*. 3.2kW solar panels and another eSmart3 charger before winter (0%)


edit:attached pic and.stl of my latest pack (#8) and mounting bracket

image_aeshwi.jpg

Man this is really helpful. thanks so much for sharing the print design. I am printing one tonight to trial for my build. it looks like it could be a real winner for me.

thanks again!
 
Man I love the cell holders. I am having another go at printing a 104 pack. wish me luck finding level on my dodgy bed surface!
 
ajw22 said:
Hanssing said:
ajw22 said:
* Finally settled ona very simple2x 2mm^2 straightsolid copper bus bar, which should haveno problem with my max 25A use.

You are aware that 50% of your cells are fused with 2x the current right?
In case of a short an outer cells will only see maximum the fuse-rating.

But an inner cells will se 2x this current before both leads blows.

Yes and no. Suppose one of theinner cells short circuits. Onone fuse we have the 1 outer cell with 100mOhm of internal resistance. On the busbar side we have 100 cells with effectively 1mOhm internal resistance. So the vast majority of the current will be flowing from thebusbar side, breaking that fuse virtuallyimmediately. So just during the briefest time the fuse requires to vaporize, more current will flow.

Perhaps of bigger concern is what might [not]follow that. That1 outer cell might not be able to supply enough current to break the fuse connecting it to the shorted cell. Thus continue todischarge and heat up that area.
But considering most manufacturers weld up to 6 cell in parallel without fuses, it's probably not a huge safety issue. But good thing I keep my system in a detached concrete bunker, just in case.



======================

Just a quickupdate for now, will post proper one with picturesin a few weeks:

Installed and running:
* 6.5kW of solar panels just for charging
* 2x 60A eSmart3 solar charger(limited to 28A ea)
* 1x 14S92P and 1x 14S104Pbatteries
* 2x 1kW GTI with limiter
Cycling happily between4.05V ~3.40V,outputting upto about 13kWh AC. Used just 40kWh from the grid this month vs 500kWhlast year.


Todo:
1. Move batteries to bolted down dedicated rack(80% done)
2. finishanother14S104P battery (70% done)
*. add another 1kW GTI
*. harvest and build more 14S104P batteries (0%)
*. 3.2kW solar panels and another eSmart3 charger before winter (0%)


edit:attached pic and.stl of my latest pack (#8) and mounting bracket

image_aeshwi.jpg

check out print. External layers perfect...first INTERNAL structure layer is SKETCHY as! high speed and under-extruding. every subsequent layer is spot on.


image_wayluw.jpg



we got there though! this file works and its away. Thanks again for the awesome share.
 
Back
Top