Raspberry Poo

Re-format the micro SD card.

Copy Raspbian to it.

Boot in RPi and set up.

Uninstall loads of unwanted stuff, and prepare to make it a read only FS.
https://hallard.me/raspberry-pi-read-only/

Set up WiFi static connection.
http://omarriott.com/aux/raspberry-pi-wifi/

Install required stuff (after finding out what it is again).
http://hertaville.com/interfacing-an-i2c-gpio-expander-mcp23017-to-the-raspberry-pi-using-c.html
plus installing libi2c-dev

Get VisualGDB connecting and building to the RPi again.
(different version of Raspbian so need to resync libs and include files)


Call it a good days work, reboot the RPi before shutting it down to backup the SD card.

Kernal panic... unhandled exception in interrupt handler... can't boot it at all...

Yet another trashed FS, this time it didn't even take a power cut, just a reboot.

<sigh>


That micro SD card has now been stabbed and thrown in the bin. Faulty card? Poor micro SD to SD card adapter? RPs just don't like micro SD cards?


Time to investigate which Arduino has 16 analog inputs, one PWM output (they all have) and another 11 digital outputs.

For now I've dug out an old 4GB proper SD card, see if that will work for more than 48 hours.

10th August 2015 Update - Software, micro controllers, solar panel incidence angles.

Been a while but I'm back working on this now.

Didn't have enough time to dedicate to this so I'm now working a 4 day week at my main job and working on this on day 5.

What's happened since my last post?


Software

Finished for now! Ready for soak testing. Incorporates a full status screen and diagnostic screen when run from a terminal. Still Todo: LCD screen status + LCD screen button control.


Raspberry Pi

Every time I have a power cut the filesystem on the RPi gets corrupted and I have to flash the SD card back to the stock image. Looking into using a read-only file system with /tmp and /var on a ramdisk - I was intending to have my software keep a log of past data but this can't happen with a read only filesystem. I could put the log on a USB stick, or create a remote log server (that could additionally host a web page, etc). Another alternative is to give up with the RPi and use an Arduino.


Power Sensing

I designed and built a little PCB that used a couple of 100A current clamps that read the power being used and produced a simple DC voltage that indicated how much "spare" power there was which would be fed into the RPi.

Designed and tested the circuit using the LT Spice software, worked great in that. Unfortunately I forgot that the current waveform isn't a nice sine wave and is as ugly as heck, which means that the premise that I used to design the circuit was wrong and it doesn't work at all.

So plan B is to use a current sensor and an AC to AC power brick transformer and an Arduino to calculate the power accurately and feed the "spare power" DC signal to the RPi.


Other thoughts

Looking at the data from the Wattson meter I'm wondering just how much charge the battery will get in a normal day. During the day when no one is in, there is approximately 500W load, white goods, fish pond pump (on a timer), Dell server PC (being tweaked to hopefully draw less power), all the little things on standby, etc.

And during a typical overcast day the solar panels generate between 200W > 700W. If I get 250W spare power it'll take 16 hours to charge the 4KWh battery.

So my guess is it'll get 50% charge on a typical overcast day, 100% on a nice sunny day, and 0% on a horrible rainy day.

Leading on from this I've been looking at making brackets + small motors + a simple driver driven from an annual timer for the solar panels (on a panel by panel basis) that can tilt the panels side to side to optimise their angle to the sun. This will create a small shaded area on each panel (apart from the leading edge), but I think the generation falloff caused by the angular error will be greater than 0% to maybe 10% shaded area.

Will probably buy a couple of small solar panels of the same technology that my main installation uses and implement the brackets (manually controlled) on these and take some measurements to see if it'll be worth doing it on the main panels.

Solar Storage System : System testing

Yesterday I finished all the wiring on the control board, the high power DC wiring, AC distribution wiring and sensor and fan control.

The only thing left to do is add the Raspberry PI and it's I/O boards and finish writing the software. I'll start on that next week, I've ordered a case for the R-Pi, ribbon cable and a few plugs and sockets to connect all the boards - annoyingly only two of the I/O boards have a hole to add a mounting post, the largest I/O board (PiFace) doesn't, and the smallest (2 in, 2 out analogue channels) is too small to have one, so I'll have to mount them in their own case and figure out some way of keeping them stable.

So in the meantime I wanted to test the assembled hardware. 9 SSR relays plus associated heatsinks, temp sensors and cooling fans, the three inverters, DC fuseboard, AC fuseboard, AC distribution block, and all the wiring and crimped connections.

I dug out nine nice old chunky switches to control the relays, fitted them to a bit of wood lying around, connected all the wiring up, used my DC PSU to provide the 12V for the relays and fans and, after a final check for short circuits, connected the battery and charger and turned the input mains power on.

I'm happy to say that everything works fine. At least to start with!

All three inverters running consumed about 45A, which is about half of what they should be drawing.

A quick check revealed that the main battery disconnect SSR was dropping about 1V, the 1KW inverter DC SSR was dropping about 0.7V, and the 500W SSRs dropping about 0.5V.

So the 1KW inverter voltage was 1.7V down, and both 500W inverters were 1.5V down. Not ideal, as the inverters are producing about 1KW between them. Without any voltage drop they produce approx 1.5KW. It'll do for now, my typical evening usage is 900W so it will cover that nicely.

The main battery disconnect relay did get nice and toasty, which tested the temp sensor and cooling fan which worked perfectly. Cut in at about 70 degrees C and turned back off at about 60 degrees C.

After running the system for 20 minutes the two smaller inverters turned their cooling fans on at an internal temp of about 45 deg C, and I know that the 1kw inverter cooling fan works when I tested it in isolation.

So, happy that everything was working properly I turned off both inverters, checked the battery charger and it's relays (both good), and was just turning everything off to finish for the evening.

The battery charger was on and charging at it's minimum of 6.6A. I turned the main battery SSR off, and the charger current sat at 6.6A. Uh? What's it charging? The battery is disconnected. Oh no it's not! The SSR has fried itself and short circuited. I was monitoring it's temperature and it only got upto about 80 degrees C, and I have checked the specs of reputable brand SSRs and they are good for 125 degrees C. Just goes to show that if you buy cheap SSRs, you need to derate their maximum current and their maximum temperature by a lot!

Not a big problem, as it was dropping 1V across it I think I will change it for a good old fashioned 300A physical relay. I didn't want any moving parts anywhere (yes, I know, the fans are moving parts!), but to get the system working at maximum output power it's looking like I'll need to change all the DC SSRs for normal relays.

If any readers know of any eBay sellers selling decent 300A relays (and 100A relays too) please do let me know.

Well, it is a prototype system so finding out parts aren't suitable is part of the game! :)

Piccy of the controller board (excuse the mess at the bottom of the board - this is all the controller wiring connected to the switches, this will all be tidied up when the R-Pi and I/O boards are installed)

Components from left to right, top to bottom:-
Top row
* DC current shunt (partially hiding behind the black box)
* DC fuse box (the black box just under the DC current shunt), 250A main fuse, 75A for 1KW inverter, 60A for each of the 500W inverters, 100A for the charger)
* Main battery SSR (fried)
* Charger DC relay
* DC ground distribution block (hey, it's huge I know, it's what I found on EBay and I couldn't be bothered to chop it down :) )
2nd row down
* DC relays for 1KW inverter
* DC relay for 500w inverter #1
* DC relay for 500w inverter #2
3rd row down
* 1kw inverter
* 500w inverter #1
* 500w inverter #2
4th row down (just below the 500w inverters)
* AC relay for 500w inverter #1

* AC relay for 500w inverter #2
* 4 x main fuses for the AC connection for the three inverters and the charger
* Main AC distribution block
5th row down
* Main 240V AC grid connection (almost off picture to bottom left)
* Charger 240V AC connection (almost off picture to bottom left)
* AC relay for 1kw inverter
* AC relay for 2.5kw charger
* Temporary manual switches used to test the other components

Solar Storage Unit starting to come together

I've got all the bits I need at last, and the external case is being built and should be here in a couple of weeks.

All components are being mounted on a spare cupboard door I had lying around which happened to be the ideal size.

Four rubber feet on the bottom of it with the signal and sense wiring to the controller on the underside, and the DC and AC high power connections on top.

All the DC solid state relays are mounted on small heatsinks with their own fan, controlled by a 70 degree C switch mounted on the heatsink.

The past couple of days I've been mounting the heatsinks, fitting the fans and done the positive DC wiring:-

The heatsinks came with one milled flat space for one SSR, but the top right one required two SSRs fitting, so I had to finish fitting a new DC motor to my CNC mill so I could use that to mill the whole of that heatsink flat so it would accept two SSRs. All things that take time!

You can see fitted to each heatsink the little temperature controlled switches, they just look like transistors. Better than the old style thermostat type switches. The fans require 12V 0.25A which these will handle with ease. I will also add a second feed to all the fans, via diodes, so that the controller can turn them all on when the overall temp in the unit case gets too high. The inverters are also getting one of these temp switches each which will feed into the controller to let it know that inverter is too hot. I also have different style temp switches that I'll (try to) fit in between the battery cells. I want to be able to monitor all parts of the system and either not use parts that have failed (or are just too hot) or shut the whole thing down.

Tomorrow I'll finish the negative DC high power wiring and then the controller wiring, and then mount the inverters and finally the AC wiring side.

This board will sit next to the battery in the case, with the charger unit suspended above this board on its own little shelf. I've also got two temperature speed controlled fans which will get mounted in the case, left and right.

I know this project has kind of grown from a few parts that were going to just be assembled on a desk, to something that monitors itself in loads of ways and will shut off if needed, so it can be put in a garage and completely forgotten about by anyone.

I just hope that at the end of it, my solar panels produce enough extra power to charge the battery in a typical day!

Richard

Inverter progress (yes, there is some!)

I have now received a bench power supply to test my grid tie inverters, and a 1Kw GTI that operates of a lower input voltage.

The new 1Kw GTI was given the once over (resolder a few cracked solder joints, replace heatsink compound with decent stuff, check for short circuits and clean out any rubbish inside it), connected to the bench PSU and sprang into life.

It starts working at approximately 24V - good, cuts out at approx 20V - good, and the MPPT algorithm starts off drawing a very low current and increases the current draw upto the maximum - ideal for use with a battery.

Connected it to grid, the LifePO4 battery via 60A fuse and a 100A current shunt, connected two meters to monitor the DC input voltage (to see what voltage drop there is in the system - a quick way to spot bad or loose connections) and the current passing through the shunt, and then turned on the grid, battery and inverter.

It ramps up the current demand over approximately 30 seconds. The maximum power generated is 850W for approx 1100W input power, so it has 77% efficiency.

No noticable heat being generated after a couple of minutes of operation, and the fan did not come on so I assume is temperature controlled.

So the new 1Kw GTI is working nicely.

I also tested one of the 500W inverters on the bench power supply. The supply can output upto 60V (at 5A) so should easily have enough voltage to satisfy the higher input voltage required by the 500W inverters.

In a previous blog I wrote about how I tried to use one of these 500W inverters. These have comprehensive diagnostic LEDs that indicate the grid voltage and frequency, whether the input voltage is too high or low, a 6 segment bargraph that shows the output power, a blue LED that blinks when the MPPT is tracking the input, and a ref fault LED. I tried connecting one to my battery and after testing the input voltage it simply lit the input voltage too low LED and the fault LED. Hence I thought they required a greater input voltage to start to work.

After connecting them to the bench power supply and increasing the voltage to approx 30V the GTI was doing exactly the same thing.

Increase the voltage even more and it then settled with the High Input Voltage + Fault LEDs lit. Which didn't appear to make much sense. Too high, too low with seemingly no middle ground?

I then tried to connect the grid side of it, just to see if that would affect anything.

Voila! It goes through the self check, then checks the input voltage check, THEN turns on the fault LED (regardless) AND either input voltage too low or too high, along with the grid voltage and frequency green LEDs, and after a couple of seconds the blue MPPT light comes on, the fault LED goes OUT and the power generation bargraph starts to work as it starts outputting power.

So the 500W inverters do work. I've no idea who designed them, but the fault light doesn't mean fault when you first turn it on, but it may do later on, after it's started working.

Input voltage required to start them is approx 24V, low input voltage when they stop working is approx 22V.

So I have two 500W inverters (can generate 400W each) and a 1kW inverter (generates 850W), so 1650W of GTI output power.


Another problem has occured to me now. A GTI inverter doesn't care what the power consumption of my house is, it'll just generate the maximum output power for all the input power it can get, and if I am not using that power, it will just be exported to the grid.

So if I was to turn on all the GTIs they would generate 1650W, I typically use 800~900W in the evening, and so approximately half of that would be exported, which kind of semi-defeats the whole purpose of storing it in the battery in the first place.

I have considered using a grid-interactive inverter (Outback Raidian for example). With one of these I would need to add a second fuse box that is powered purely by the Outback, and connect my low demand loads to it (no cookers or tumble dryers, etc). I can then connect my solar PV GTI to the Outbacks *output*, and in the event of a loss of grid power the Outback disconnect from the grid and uses the battery to supply power to my house loads and to also keep the solar GTI online and feeding solar power to the house loads. (a slight problem with this is when in this mode, if the solar PV is generating more power than I am using, it charges the battery in an uncontrolled way which can easily fry the battery)

For the moment I will carry on with the three GTIs I have. As I have three of them, 500W, 500W and 1000W, I can control them individually and roughly match my power demands. Normal power demand ni the evening without the TV and AV system on is approx 500W, so one 500W GTI will supply that nicely with minimal loss (where 'loss' = the GTI output being more than the load and the excess being exported). With the TV and AV system on load is approx 800~900W, so the 1KW GTI will supply that with 100~200W being exported.

I will have to experiment with the hysterisis points around the 500, 1000 and 1500W points, and the time before turning another inverter on, and time before turning an inverter off, taking into account an inverter takes about 30 seconds before it's generating full output power, so loads like kettles (on for maybe 2 to 3 minutes), microwaves (on for 1 or 2 minutes) do not turn extra inverters on which will be of very little use before the extra load goes away.




Richard

Inverters (again...)

Status update:-

PowerJack : No reply to enquiries in to what CPU they are using in this years models.

OpenEnergyMonitor forum : No replies so far to my post asking for model numbers of Powerjack inverters that are known to have the ATMega CPU.

The Chrome browser does do a nice job of automatically converting German to English - going the other way I'm guessing is laughable but I'll join the German forum and post there anyway, at the very least I'll make a couple of members laugh at the translation!

I've ordered a dual bench power supply (0 - 60V @ 5A or 0 - 30V @ 10A) which will let me play about with various input voltages for the inverters I've got and see where they "wake up" and what the current consumption waveform is like - I have heard people have discovered their MPPT inverter when connected to a battery source, wildly varies the current drawn between almost nothing and double their rated input current - not at all good for longevity.

(I have various power supplies, 5V - 17V @ 35A, 24V @ 45A, 0V - 14V @ 1A, but nothing that goes upto the 55V these inverter can require, and nothing with current limiting, save a selection of fuses!)


I've traced out the DC voltage sensing circuitry in these 500W inverters and now know what is needed to get that part working from 26V, so when the power supply arrives I'll do that and test one. What the rest of the inverter does when working off a low voltage totally depends upon the design and CPU code.


Other alternatives I've been looking at are large online UPS modules - the problem is that any decent UPS uses a DC battery voltage of 72V, some even more, so they're unusable for my project.


Also looking into just getting an 8WK non grid tie inverter, adding a second fuse board and swapping over all lighting and all the sockets (apart from kitchen, utility room, garage, workshop - anywhere that has a high power load) and a switch over relay to switch back to grid power when the battery goes flat - the switch over does concern me, I suspect if the inverters waveform was 180 degrees out from the grid there could be an instantaneous change from (say) +270V to -270V (rough absolute max voltages) which would, well, I'm not exactly sure what it would do. Generate a short burst of high frequency noise for one. And I imagine any inductors will try to maintain the +270V and generate a lot of back EMF.

Saying that, this is how offline UPSs work, a simple relay without any inverter waveform synchronising.

There is also the problem that may arise if some unsuspecting person plugs a high load into the inverter powered circuit. I can get an 8WK inverter with a 30 sec peak of 16KW which is way more than one device (connected via a 13A slow blow fuse) should use, but instantaneous switch on currents can be very high, 10 times rated power isn't uncommon, so 30KW for a few mS.


Hence I will continue to pursue the GTI solution for now. This new PSU can supply enough voltage to "turn on" any GTI, so I can check my modifications and also connect my scope and see what the voltage and current draw is like. Also the new 1KW GTI should be hear next week, and that should be fine for working from my battery.


I've also read a lot of (valid) concerns from people about these inverters that just plug into a wall socket, and taking their concerns into account in the final system I'll have to modify the GTIs to remove any output sockets and make sure they're hard wired with proper earths and protection in place, enclose all components in a decent ventilated box / case, and also locate the system next to the fuse board and put it on its own spur. All common sense really.

GTI Inverter headaches

Just a quick post to update readers on the avenues I'm exploring about this problem.

One possible solution I have found is that some Chinese inverters use an ATMega controller CPU. There are a group of great guys in Germany who have discovered these, interfaced to the CPU (this type of CPU is easy to reprogram) and written their own version of the firmware that allows the GTI to be reliably used with a battery.

I found this post here that refers to the German forum and details the code a little
OpenEnergyMonitor.org
http://openenergymonitor.org/emon/node/1658#comment-21742

The German forum is dasWindrad.de, here is the linked post
http://www.daswindrad.de/forum/viewtopic.php?f=19&t=860&sid=b206245d4f6cf8c43dd314799e57875a&start=80

which could quite possibly be the answer to my problems if I spoke German, which I don't.

The GTIs which they say use these ATMega CPUs are made by PowerJack.

A blog from what appears to be a PowerJack employee (in the design dept?) is here
http://powerjack888.blogspot.co.uk/2014/05/lf-psw-power-inverter-from-power-jack_26.html

And I've send a message to the fellow, Chang Jack, asking if he can point out which 2014 products use the ATMega CPU.

Another possibility is REUK:-
http://www.reuk.co.uk/Grid-Tie-Inverters.htm

This chap built his own GTI and has made all the designs and source code available for anyone to build their own GTI, and obviously that will be easy to disable MPPT and optimise it to work off a 24V battery, but it's not exactly scalable if I want 10 inverters.

I will keep this blog updated when / if I hear back from my enquiries.

Oh, one last thing I intend to try with the little 500W inverters I have (that don't recognise the 26V from the battery as enough voltage to start working) is to adjust their voltage divider (there's a convenient potentiometer) to see if I can persuade them to at least start, and then see how their MPPT behaves with a low resistance (the battery) supply.