Project Daft Idea

Nathan

New member
Joined
Dec 5, 2016
Messages
28
:D :D :D

Firstly, Welcome to my project thread, where I will be documenting my Long (And very slow) progress through to building a powerwall Incrementally in values up to around 10-20Kwh.

I am starting from literally no experience in this, apart from watching all 172 videos from Peeete

I have a lot of enthusiasm, not a lot of money, and a few skills I will be throwing in the mix as the project develops into what I hope to be a really good end product, using a data driven approach. The more data I collate the better. As a professional software developer, im sure my skills will be improved by using them in different ways as well.

Will soon be showing the projess in the form of pictures and screen shots, got my initial 50 cells here to work with and map and develop the software over christmas to ensure we get a good end product. once I have firm designs it is my intention to release all hardware designs/software on a free for non-commercial license.

So now onto the project goals.

1. Build a power generation/storage facility that can negate the power usage of my business ~10kwh per day.
2. Put in place systems to evaluate and monitor individual 18650 cells, providing discharge curves, ESR, Initial voltage when recovered, Voltage at each top up charge.
3. Ensure that batteries are always operated within there safe limits
4. Minimise the amount of labour required in testing cells/gathering cell data.

Onto Point 4, I have already built a arduino based charger/discharger which charges at 500mah via a tp4056 and then discharges a single cell using two resitors. Providing the data over the USB lead. This has severe limitiations in the fact that it will only work on a single cell, and the charge limit is very low due to the fact it is charged over USB. Therefore I am upgrading this design to utilise a ATX PSU, which can provide 49 Amps via the 5V/12V lines and convert them both to a single 5A line running 30 Amps, which will be used to charge 30 Cells at once.

Once charged, the system then switches to discharge mode and outputs the data over serial, currently into a terminal program, but this will be altered to use a C based program which will log it to a MySQL database, from where I can produce graphs, each cell will be checked for power consumption ever 1 second. giving a good scope of data of how the cell is operated.
 
Oh there are a lot more 18650 videos than just the ones Pete made, I have quite a few youtube subscriptions.
 
Do you really need to track that much data on individual cells? How are you going to tell which one is which? I suppose you could barcode each one, but once they're in the pack all running in parallel, do the individual details matter anymore? Don't get me wrong, what you're doing and the data you're logging seems very cool. I'm just questioning the usefulness in relation to the amount of extra work/expense it is generating.
 
station240 said:
Oh there are a lot more 18650 videos than just the ones Pete made, I have quite a few youtube subscriptions.

I know, im currently watching through Paul Kennetts channel as well, whats your channel and ill start watching those as well :D
 
mike said:
Do you really need to track that much data on individual cells? How are you going to tell which one is which? I suppose you could barcode each one, but once they're in the pack all running in parallel, do the individual details matter anymore?

No, Thats one reason ive called it Project Daft Idea - and im a bit OCD :D :D

Got it in one, every cell will be barcoded for identification. And yes during day to day running you can say that individual cells don't matter as there in parallel. However I am planning for future failures and ways I can try and pre-empt using bad cells further in the future. Upon failure, identifying a poor cell will be quite easy, firstly sticking thermally sensitive paper (sticky labels with heat printed barcodes on ;) ) will help you identify which cell caused the failure (the label will be darkened by the heat).

From the identification I will be able to bring up its initial discharge trace and compare it to the averages and get the computer to compute other cells with similar patterns.

Also further from this, each pack will be barcoded, and all cells registered to the pack - a process that will take a minute or so with the barcode scanners I already have per pack, meaning if the label gets ruined, I will be able to see from the other cells assigned to the pack, which one this cell is.

mike said:
Don't get me wrong, what you're doing and the data you're logging seems very cool. I'm just questioning the usefulness in relation to the amount of extra work/expense it is generating.

Thanks, as Ive called it daft idea, sometimes things are done just to see if it is possible rather than for sensibilities sake.

For every battery charger/discharger circuit, my current plans are to have 4 per atmega328p, with 4 TP4052 and 2 ceramic resistors, a crystal and a couple of capacitors. In total per unit the cost is around 8 in components, 5 for the PCB. There will be a total of 6 of these and they will handle all charging/discharging - that comes in at 13 per unit, or 78, of which I have dozens of Atmega328's lieing around as I buy them in bulk for other projects I do for fun. if it comes in at less than 100 for the charger/dischargers I wont be disproportionate with anyone else out there. As for the work, developing of the system is done in my R & D time, done to learn and better my coding abilities (As a self employed programmer, I allow 1 hour per day for R & D), any coding that pushes my limits in terms of memory/resource management is useful for my day to day business and so If I wasnt doing this, I would be doing something else and as such, the time costs can be offset.


Ok, few statistics after ive opened my first 90 Cells and initial voltage check.

My cut off for assuming a cell turns up in good condition, is it is above 3.5V, and below 4.2V. of which 46 out of 90 meet this target. A further 30 are in the 2-3.5V range, leaving 14 below 2V

Of those 14 below the 2V range, 1 has a rusty positive end and voltage in the milivolts range, and the voltage on another starts at 4.6V and rapidly sinks to .6V. Both are simply being put in a scrap bin for recycling.
 
Keeping track of every cell's spec/original source is worth doing IMHO.
If you get cells that look/test fine, but fail later on, you can see if they all came from the same laptop battery.
If you note what pack they went into, you can decide if you want to bin all the other cells from that laptop battery.

Nathan said:
station240 said:
Oh there are a lot more 18650 videos than just the ones Pete made, I have quite a few youtube subscriptions.

I know, im currently watching through Paul Kennetts channel as well, whats your channel and ill start watching those as well :D

I don't have a youtube channel, as my internet upload speed is 0.3Mb/s. It's not just crap, it's overpriced crap.
Look up the channel "UK DIY TESLA POWERWALL" instead.
 
station240 said:
Keeping track of every cell's spec/original source is worth doing IMHO.
If you get cells that look/test fine, but fail later on, you can see if they all came from the same laptop battery.
If you note what pack they went into, you can decide if you want to bin all the other cells from that laptop battery.

Yep, I am tracked back to the model of the battery and date code if applicable on the battery, as well as also noting down each cells branding and model number.

THe idea is if i find that some insertcrappybrand B18650-uttercrap keep pulling my system down, I will be able to track the others of the same cell model and remove them from the system completely, in an attempt to make second hand rubbish.

Also the data points will eventually be able to help me predict the duff batteries from there initial cycles, and remove some of the randomness element from using used stock.


station240 said:
I don't have a youtube channel, as my internet upload speed is 0.3Mb/s. It's not just crap, it's overpriced crap.
Look up the channel "UK DIY TESLA POWERWALL" instead.

I will do - Thanks for the heads up!


As for that internet speed, if mine was that slow id probably cry real tears
 
Phase one for the first 29 Batteries completed. There is a long way to go, but the early signs are great

Some statistics based upon a voltage check only, out of a total of 173 Cells coming from a sample of 29 battery packs.

126 Cells > 3.5V
28 Cells > 2.0V but less than 3.5V
10 Rows > 1V
9 Rows less than 1V.

Out of the 173 Cells, 2 are scrap due to physical damage, Both are LR1865SE from different battery packs, and a further is failed due to falling rate voltage as mentioned above.

I have another 19 Battery Packs on there way to me, this should get me to a fairly nice place in phase one, testing of my electronics.

Whilst these turn up, ive got to expand my design from working one battery per arduino, to working 4 per arduino uno's, finalize the PCB layout and get them ordered for delivery in the new year when the discharge/recharge process will start
 
Ok, so Prototype 1 is currently charging and discharging its first full 18650, pumping data into CSV's for our other system to process, evenutally will get it built into a PCB once its thoroughly tested with a few dozen 18650's.

So whats on the board, Songle SRD-05VDC-SL-C relay switching the battery between a charge and discharge mode, a couple of 10Watt resistors, one as a shunt resistor to measure the voltage drop, and one acting as a load, drawing around 1amp from the batteries in discharge. A IRF3205 Mosfet to switch on the discharge, as the resistors are also used to calculate the Internal resistance.

A TP4056 IC charging the battery at 1 amp, well half an amp at the minute as its only powered over USB, All of this is controlled on this demo board by a Atmega328P-PU running the Arduino bootloader and custom code.

The 3 LED's are for Charge (red), Discharge(Amber) and Cell Complete (Green), the code is currently a bit of a mess but all components work, just not been through a complete cycle, which should be tomorrow.

Once ive run a few 18650's through it, I will be building up the schematic based around a ATMega2560, as it will allow me to have 16 Analogue inputs, 54 digital inputs, allowing me to run 8 18650's and data log them.

One pic of the prototype:

image_sncnva.jpg


as far as costs go, the only unknown cost is the PC, the per battery costs is a shade over 1.20, with overall arduino costs of 9 in terms of IC's, say a 3 in other sundries.

This means, to charge 24 Batteries simultaneously, gather data on them, will come in at 38 plus the cost of the PCB's to be printed, Using an external supplier like Osh Park, I expect if i can keep the board size reasonable id expect this to be around 20 for 3 boards.

If this goes to plan, I may even go all out and get a second batch done if i decide to grow to 10Kwh.
 
I get what your doing just don't understand it - looks cool tho :)
 
hbpowerwall said:
I get what your doing just don't understand it - looks cool tho :)

Thanks,

I get what im doing, understand the end goal as to why im doing it, just dont fully understand why I am doing it.

I mean ive spent 100's on used laptop batteries, creating 100's of hours additional work for me, but Why?

We have very good electricity supply, that is of reasonable cost (compared to investing in solar panels, a few kwh of batteries etc) and my time could be better spent on other projects..

But theres something in me thats saying this old rubbish shouldnt be just discarded, should be used, yet some of it is rubbish and should be recycled.

Theres also the fact im pretty sure I am OCD about things, everything has to be either perfect or cobbled together, prototypes are cobbled together and 3 or 4 components on the board have been disconnected and left - but on the PCB's im putting together it will be as near perfect as I can get it. I dont want to spend a long time building up a pack for a single cell to give me doubts I can run it.

I will be using a small pack design, 16 cells per pack with a target minimum mAh of 2200 per cell, I will be needing 224 cells meeting this criteria, as im also thinking of going straight in at 48V if I can find a reasonably priced inverter, would mean 14S. This would give an initial packsize of around 1.7 Kwh.

Are my maths right? This seams like a large pack for such a small number of cells..
 
note : i suck at math :) lol

Understanding why I've gone this far illudes me also...
 
Would a PIP4048 work properly with a nominal voltage of 48.1, dropping down to 45.5, id thought it would be ideal to keep it at 48V minimum...
 
Firstly.. Merry Christmas and a Happy New Year

Ok, now back to project daft idea.

Over christmas I have designed the relevent circuits, split it up to make it simpler (and cheaper) to make and sent for the first batch of boards to control the LED status lighting for each battery, made up the chargeshelf and now im waiting for our friends in china/malasia/America to get the goods to me to finish assembly..

A few Pics..

First the underside of the shelf, never been a fan of wires hanging down, and with 24 RGB (4 leads each) LED's drilled into the edge, that would leave a lot of wiring, well we have cut out a half the material and will fit 6 custom PCB Arduino's to manage the LED's (4 LED's each)

image_zejcuv.jpg


Using a spare PCB from another project to make sure the leads fit on top for the LED's

image_ghdgiz.jpg


How the top will be laid out, can just see the LED's popping out at the end of the shelf.

image_njesgk.jpg


A final look where im up to, the data will be logged to the PC underneath once these have all been finished, and when finished every wire will be neatly tidied away as i can be a little OCD.

image_dzxfmt.jpg
 
Thought I would throw my hat in the ring, Here is my take on the cheapest way i could come up with to have at least 5 individualcells being controlled by 1 328p, and i may have missed something, but i believe its fairly bulletproof, (e.g. battery installed backwards), its intended to charge, discharge, test cell ESR, andyou could check self discharge if it took your fancy.

You would have up to5 of these circuits, with an ADC dedicated to each cell voltage line, and likely a 6 or 8 way analog switch to loop through the current readings (you only need to integrate them a few times a second to be accurate)

Maximum current is ~1A, however the charge current will max out at 0.86A or there about when the cell is near full, (5-4.1 =...). This limit is only due to the size of the shunt resistors used in each loop, this was more to ensure the power dissipated in the silicon device was small so cheaper mosfets could be used. and any kind of fault current would be well within the limits of the circuit. pretty much a 500mA unit that has some extra.

When you enter discharge mode, The body diode of the charge side circuit will allow current to flow and complete the loop, this will then cause a negative voltage to appear on the charge circuits feedback trace and slam the mosfet fully on, The op amps specced can happily run up to -0.3V on any input without weird behavior, and the schottky diodes and resistor networkon the feedback traces ensure it can never exceed this rating. from then onthe discharge side circuit will regulate the current.

At the top, there are 2 back to back P Channel Mosfets, this was done on purpose so that the body diodes cannot conduct and do damage if someone reversed the battery in the holder,

And to wrap up my only other notes would be, for the current set points, you will probably want to add an RC filter, (you have 6 PWM outputs), The mosfets should be rated for at least 5A, and the 4.7M resistors arethere to add some hold-off so the circuit doesn't start conducting until your asking for more than the input offset of the op amps could add up to.

image_fckepq.jpg
 
Whats the status on Project Draft?
 
No idea on the op, but a little rethinking on my end means you could actually have a fair bit more while using less peripherals if you wished,

You can get away with 3-4 bits of a shift register for your setpoints if you wished, to set say between 1.6A and Off in 8 or 16 steps just using resistors. (Ladder DAC), then use the last bit for the the charge discharge mode, this lets you replace the 2 analog mux's for 1 shift register.

Your then left with 5 wires, but it can be daisy chained much further, Serial In, S_Clock, S_Latch, and your voltage and current outputs, as the set point should not need to change very often,

Depending on how you arranged the analog voltages you could have a very large number of these connected all working independantly, as the current loops will keep on doing there thing regardlessly so long as the cell ESR isnt so high that it saturates

As the QFP package has 6 ADC's with I2C comms, or 8 without, and using SPI, you could well make up banks of 4, and have the brain box analog mux through them.

equally would involve reducing the shunt resistor to say 0.47 Ohm just to give it some more headroom for high esr cells.
 
Sorry I have been busy with Reorganising and upgrading our office for the past few weeks to try and create more space for stock whilst also awaiting IC's from China.

I love your technical knowledge Rerouter - Your solutiong looks more elegant than my solution, but with your experience id expect that! Your solution will probably work out a lot cheaper as well. I will be implementing your solution as a test case with hope to learn more!

I have all the lighting sorted for the charge/discharge state, I will be using INA219's for the current measurement, TP4056 for the charge control as it is a fairly cheap way to control the charge cycle and has a relatively easy method to get the charge state back to the atmega328P, This left me with a challenge regarding needing about 6-8 digital pins and so MCP23017 Port expanders are being used to resolve this issue.

On my demo circuit it appears to be working, if that will scale is still to be shown. Either way it means I will be using 2 328P's for the 24 cells and 2 USB ports on the PC to interface with the software

Will et some pics up as soon as there is something contrete to show everyone
 
Back
Top