Author Topic: Modelling solar power using supercapacitor for battery bank protection  (Read 1047 times)

0 Members and 1 Guest are viewing this topic.

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
On http://www.anotherpower.com/board/index.php/topic,902.msg10541.html#msg10541 I described and showed the changeover switch for switching circuits in the house from grid to off-grid on a timer so as to take advantage of cheap night time electricity, and to adjust the off-grid use in winter when the solar energy in the UK simply isn't enough.

I've had problems with my battery bank tending to take too much stored energy out of it in the autumn leaving it too depleted in the winter with danger of sulphation of the batteries.

Accordingly I've been wanting to model the amount of solar power being generated so as then to control switching back to the grid so as not to take out more energy than is put in.

There are now available 500uF 2.7V supercapacitors on ebay and I think that people are having fun putting them into banks to start car engines instead of batteries. I have made up a 24V bank of 12 of these to put on the terminals of my inverter to reduce ripple but have found that there was no reduction at all and perhaps its not impossible that these small size capacitors have too much inductance to smooth ripple but are actually better suited simply to storing energy.

My off grid system has about 8kW of solar panels fitted to a barn roof, some facing east and others facing south with a fence made of 8 broken panels (cheap) facing southwest at a near vertical angle to catch late afternoon sun. This feeds a battery bank and an inverter, and then is connected to the house through a 300m or so cable. With the cable we laid a couple of CAT5 cables to provide signalling functions.

I seem to have troubles with the CAT5 cables at points from time to time nibbled by rodents or suffering for other reasons so that of the 16 available wires only 6 or so :-( are working.

Through the available signal cores, in the house, I'm able to monitor battery bank voltage, one voltage of one set of solar panels before entering the controller, and the voltage of another set also. I put LED lamps between the negative of the monitored voltage of the panels and the negative of the battery signal - the solar controllers are common positive. This means that the lamps flash according to the off periods of the PWM controllers and stay on when the batteries are at full voltage so that the controllers have cut off, letting the solar panel voltages raise to their open circuit value.

Luckily there's one more pair of wires in my CAT5 cables that are working that are available.

As a start of a control circuit, at the barn on the roof I mounted a small 12V 5Watt panel. I was experimenting using that to work two electromagnetic relays in the barn, intended for control use. From that solar panel feeding the relays, providing a load, the voltage is fed into the CAT5 pair of wires. In the house, this voltage has been feeding a voltmeter and a pair of 12V led lamps in series which light when solar energy is available in goodly excess.

In veiled sun the resulting voltage at the house from this monitoring panel has been around 9V and in full sun the voltage has gone up to 18. This accords with my real system where in veiled sun around 80amps is going into the battery bank and around 180amps in full sun - so between 2 and around 4 1/2 kW input.

But much of the time in England this winter has been dreadfully cloudy and at times I've been lucky to have 10-15amps going into the batteries.

Having the monitoring voltage coming into the house I had the idea to feed it to a supercapacitor as an analog to the battery bank.

Whilst much can be done with electronics or arduino control, to be frank, in 25 years time people won't understand component level electronics nor arduino code. So I have wanted to do things in a really dumb basic way that is so simple that anyone in the future will be able to understand it.

Protecting the supercapacitor from exceeding 2.7V - OK zeners are fine but LEDs are pretty :-) A red LED fires at around 1.5V and a green starts to fire at around 2.2V and blue at around 2.5V. So that's a good start for protection. With the quantity of current available from my 5W solar panel, the voltage on the green LED was approaching 2.7 volts and even 2.75 (ouch - but I assume a small margin might be tolerated). So I'm going to be using 3 or 4 green and blue LEDs in parallel as well as a red LED with a small resistor, ideally keeping the supercapacitor below 2.5V.

On my first test, it took a day of 50% veiled sunshine to charge the supercapacitor, the 5W panel powering the two relays in the barn and feeding the capacitor through the resistance of the 300m cable. As a guide, from memory the pair resistance of that cable is around 180ohms.

I was worried that the relays on the barn end of the cable would discharge the supercapacitor over night, but on account of the low voltage, low current resulting meant that the voltage on the capacitor went down only to 1.5V and this is a promising start.

On the next day, I've found that the 2-2.5V on the capacitor is enough to fire a 6V relay, and that on the low voltage, the current is small and the discharge therefore is still preserved in terms of being a long time. It will be interesting in due course to find out at what voltage the relay drops out.

But certainly here I have the workings of a system that integrates the solar power input into my battery bank and can fire a relay all the time that the solar power is a healthy amount. My next task is to model the analog of the power consumption.

I'm proposing to use a current transformer on my live wire from the off grid supply, to put it through a full wave diode bridge to generate a DC voltage, and to apply that reversed through a resistor across the capacitor so as to discharge it proportionately to the power use.

If it's possible to model the balance of power in and power out reasonably well then this system should be invaluable in preventing progressively taking out more power than is being supplied from the solar panels during Autumn as we head into Winter and I then need to shut it down and take next to nothing for at least a couple of months.

Best wishes

OTW

Offline Solarnewbee

  • Full Member
  • ***
  • Posts: 74
  • Karma: +0/-0
Hey the Off

2 things came to mind after the numbness went away,(actually most of it made sense)

1. rodents:having dealt with lots of chewed wires I learned from a seismologist dealing with the problem on his equipment on the tundra of Alaska that if you mix cayenne pepper powder and vaseline and daub it liberally on the wires it deters chewing quite well. works great on joint pain as well.

2. You said Arduino! Bought a Mega and an Uno R3, when I learn the code I am wanting to use Arduino modules to monitor temps, voltages and perform actions automatically. My PJ 8k inverter is being Ozified little by little and I have seen through much of the discussion that much could be done autonomously to any system to prevent outages due to inverter and battery problems.

just thoughts (or voices)

Best Regards!
SN

Any day above ground is a day for potential mishaps

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
Initial experiments are promising.

A 5W panel is loaded by 24V relay with coil of 150ohms and another of 300ohms. I use these to switch different things at different solar radiation levels. These provide a load so that we're getting a voltage to measure which bears relation to power being generated.

From there there is a 140 ohm cable down 300m of wire to the house. There we have a voltmeter measuring the voltage of the panel at the barn. Normally around 16V in full sunlight.

This feeds a 500F supercapacitor through 1000ohm resistor.

The supercapacitor is protecected from reaching 2.7V by an orange LED with a Shottky diode in series, so that combined they conduct at around 2.4V

A 3000ohm meter is across the supercapacitor.

Now the problem is to drain the supercapacitor proportionally to our usage.

A clamp on current transformer rated at 50mA per 100Amps lights an LED nicely. Maximum current drawn on the system is 25amps so the maximum current through the LED will be around 12mA

The consumption in the house is roughly constant at 750W to 1.2kW with bursts of 2-4kW but totalling a similar amount during the day. So provided we get the ballpark range of load on the supercapictor right then exact linearity is a luxury rather than a necessity.

The GJ5516 LDRs are just the right diameter to couple with a green LED. File the LED lens flat and stick on with superglue.

Attached to the current transformer, 750W gives a resistance from memory at around 1200ohms and the 3kW consumption decreases the resistance to around 300ohms.

Loading the capacitor with this should give me something to play with, varying the 1000ohm feed or doubling up on the LED/LDRs in series or in parallel giving a guide as to daily power in vs daily consumption.

Has anyone any better ideas of achieving the goal?

Best wishes

OTW

Offline Solarnewbee

  • Full Member
  • ***
  • Posts: 74
  • Karma: +0/-0
Howdy Off!

These kids that build enormous sound systems in their cars have been using 1 farad low voltage electrolytic caps to reduce noise from the alternator (gen) and magnetos. Would such a cap reduce the ripple? I always thought the lower the farad the more noise reduction and ripple is noise right?

SN
SN

Any day above ground is a day for potential mishaps

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
Oh - I was going to write about this on another thread. Apparently whilst pulses are good for desulphating, batteries don't like ripple. I have tried a bank of 12 500F capacitors on my 24V system and suspect that they are so compact that their inductance is too high - and as a result, don't reduce ripple at all.

But I bought a bank of 1F 16V car audio capacitors to try to smooth out a conventional set of solar panels, of course without success, and so instead put them in pairs for 24V so with 8 pairs I've got 4F on the terminals of my PJ inverter. This removes about 1/3rd of the ripple. Does anyone know whether it's important to remove discharge ripple?

Best wishes

OTW

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
The supply aspect of the modelling as described takes the capacitor charge from 1.5V to 2.5V quite happily in a half day of sunshine so this is useful as an indicator to start with, using a 1000ohm resistor into the capacitor.

The problem is modelling consumption.

Using the current transformer as described attached to a green LED I found the GL5516 to be logarithmic and I really didn't think I'd get away from that. It registered around 750ohms when the consumption was around 1kW and 350 ohms at 3kW. Trying an ORP12 I've found it giving much better range and possible linearity in the range under consideration so that it gives around 1200ohms at 1kW and 400 ohms at 3kW.

The smaller dynamic range of the GL5516 in this situation during full sun all day meant that the capacitor voltage struggled to rise 1/2 volt in a full day of sun.

It's necessary to put two LEDs back to back in loading the current transformer so as to take both halves of the wave. Having both halves of the wave to play with means that to vary the response rate one can have two ORP12s in series or in parallel accordingly. This can enable either time to be proportioned or rate of consumption vs supply.

I may have two indicators, one based on a day to day basis to warn when consumption is more than available power on a daily basis and another on a four day basis.

For those running battery systems storing enough power for excess depletion to be unnoticed in winter this simple modelling should be both reliable and essential. My experience over the past two winters has been rather like driving a car without a fuel gauge.

Best wishes

OTW

Offline lighthunter

  • Full Member
  • ***
  • Posts: 154
  • Karma: +8/-0
I havent read all of what you wrote but it appears that you want a bettersystem to indicate charge level or state of charge (SOC).  I could be wrong but i think there are simpler ways to go about it.

With a large bank, (lead acid type) voltage alone can give good performance. I actually do this even now using LFP batteries. You just need a high resolution voltage monitor with time delay ability. My setup uses a plc type controller(idec smart relay) about $50 on ebay. It has analog inputs with 0-1000 digital points to indicate level on input. It can be programmed easily from the keypad on the front of device. The program can be set to switch an output on or off depending on the bank voltage with a few second time delay to avoid sensing load starting. This works well and is cheap and easily adjusted.

Another option is a state of charge meter. Several are available some better than others. Ive read some RV people love these they even tell you when to equalize or do maintenance charge.

If you want a way to monitor usage then the coulomb counter is very helpful. Its a hardware circuit that outputs a frequency and polarity indicator in proportion to amps and direction. You couple that output with a counter and tabulate the running total of current or (coulombs) assumingnyou know quantity you battery can contain.  Thats the trouble with modeling or coulomb counting , they have to be corrected for error with battery voltage frequently or they arent accurate. Even some of the best circuits struggle with accuracy (like cell phone battery level indicators). It always comes back to the actual voltage of a cell to give you the facts. Voltage cant tell you how long you can run but it can tell you "its ok to run now" or "its time to stop, battery danger "   Any good state of charge indicator must use voltage correction or they just cant be trusted.

Sounds like you have some great ideas going, I just wanted to add to the options picture. I have in the past had a real problem with doing things the hard way or re-inventing the wheel. There are always 100's of ways to do things and never only one right one. What works best for you is the right one.

Cheers
LH


LH

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
Hi!

Thanks

The system is actually coming along really well and is essentially easy.

It's essentially a small solar panel feeding current into a capacitor through a resistor. LEDs prevent the voltage going above 2.2V and cooking the capacitor. Nothing could be simpler.

We then have leds driven by a current transformer which couple with ORP12 LDRs. This provides a proportional relationship between power consumed, resistance and therefore current drained from the capacitor. So nothing could be simpler.

Of course the current drained through the LDRs when the capacitor is at 2V is four times that at 1V. So I'm using two LDRs in series at 2V and using a voltage comparator to switch a parallel pair at 1V and another LDR at 1/2V so as to give some degree of linearity in terms of the drain on the capacitor.

The problem with monitoring batteries is that of declining performance and for that reason having a measure of solar power in, and electrical power drained gives a helpful indication of the balance point and ensure in the autumn period of seasonal declining input that one's not draining the system output beyond the available input.

Best wishes

OTW

Offline rossw

  • Senior Moderator
  • Hero Member
  • *******
  • Posts: 781
  • Karma: +27/-0
  • Grumpy-old-Unix-Admin
We then have leds driven by a current transformer which couple with ORP12 LDRs. This provides a proportional relationship between power consumed, resistance and therefore current drained from the capacitor. So nothing could be simpler.

You can still get ORP12? Cripes.
Seriously, those things have a major thermal response too. Not sure its going to be anywhere near linear enough over the likely extremes of temperature to be reliable.


Quote
Of course the current drained through the LDRs when the capacitor is at 2V is four times that at 1V. So I'm using two LDRs in series at 2V and using a voltage comparator to switch a parallel pair at 1V and another LDR at 1/2V so as to give some degree of linearity in terms of the drain on the capacitor.

So step-changes in characteristics based on some arbitary threshold. Not sure that's ideal either.
I'd be looking more towards a constant-current discharge rather than the LDR arrangement - or in this case, a voltage-to-current type circuit - with the voltage being that from your sensing circuit, driving a "constant current" drain - so the drain should be largely independent of the cap voltage, but entirely determined by the sense voltage.

Quote
The problem with monitoring batteries is that of declining performance and for that reason having a measure of solar power in, and electrical power drained gives a helpful indication of the balance point and ensure in the autumn period of seasonal declining input that one's not draining the system output beyond the available input.

Given all the variables, if's and but's, I think I'd use a tiny little processor - like say, a picaxe, monitor *POWER* in and *POWER* out (which will require taking both volts and amps into account, constantly), then use a picaxe output to drive say, a radio-control servomotor or a meter or LCD panel to display SoC, along with some sort of "reset" when it sees high enough voltage for long enough...

It gives you the extra flexibility then to do some of your conditionals in code, and outputs to do "stuff"...

Offline off the wall

  • Full Member
  • ***
  • Posts: 73
  • Karma: +3/-0
Yes - I understand. But in 10 year's time no-one will happen to remember PICAXE code or whatever it is. I'm dealing with an electronic organ at the moment working on M114 chips and requiring an ATARI   8) computer working on floppies   ::) to program them.  :'(

Essentially exact linearity isn't at all required in this application. I simply need a cutoff point which can be found quite easily by observation. Drain by the household is vaguely constant on an averaged daily basis. In the summer provided we have 1 sunny day then four dull days won't drain the batteries catastrophically. So taking a cutoff point over four days, for instance, means that when that point is reached in the autumn, one can reduce power consumed (I have three ringmains switchable between grid and the battery bank) or switch to grid entirely as I have done for a couple of years between November and the end of February/March. But we've had some really grey winters without many spurts of sunshine.

Essentially when one is looking for a cutoff decision point,  non-linearities are irrelevant as one's not looking for tracking between a pair of curves but seeking that point of intersection between the two graphs.

Analog modelling in this way is fit for purpose, uses technology that will be understood in 25 years time, and is reliable.

Having bought a 2V meter rather than a 3V meter, it's not necessary to drive the supercapacitor near to its 2.7V maximum, and with 1000ohms limiting the input current to the system, 6 green LEDS as Zeners across the capacitor provide a good well defined 2.2V limit to the capacitor voltage. In fact one can get a packet of RGB LEDs very cheaply so that by using a Scottky diode as used in solar panel construction with a 0.4V or so voltage drop rather than standard 0.7V of a silicone rectifier one can use the red LEDs in the RGB packages equalising the red forward voltage drop to that of the green.

The only "electronics" necessary then are voltage comparators and these are available as quad boards such as http://www.ebay.co.uk/itm/DC-12V-4-Channel-Voltage-Comparator-Stable-LM393-Comparator-Module-/131781365781 which can be used to switch in the additional parallel LDRs as the voltage drops from 2 to 1.5 and 1.5 to 1 and 1 to 1/2 and the fourth to signal the mains changeover switch to revert to grid.

By using such a simple board and keeping the modelling analogue, this is a system that anyone can put together without expert knowledge or computer processing expertise. Furthermore by avoiding measurement of the DC input and output (rather complicated in my case in any event with multiple solar controllers from multiple groups of solar panels feeding multiple sub-banks of batteries) one's avoiding introducing extra connexions with associated risks at high current. Playing with 250 amps on a 24V system requires the cleanest of interconnexions and checking for heat on joints and any deterioration of the connexion after heating and cooling. It's for this reason that I have avoided DC measuring, making mesurement of AC output the logical preference with a non invasive clip-on current transformer.

ORP12s are easily available http://www.ebay.co.uk/itm/ORP12-Photo-Cell-Sensor-Opto-Light-Dependent-Resistor-/121596833064 and don't go wrong. Temperature dependence - what's that? Living in a stone house with walls 12 to 18 inches thickwe're looking at a range of 6 to 20 degrees inside the house winter to summer.

Best wishes

OTW