Scratch built solar PV divertor

[MikeSharp01]

Thought I would dig about a bit more on this topic - from an academic point of view at least. Found this link (https://learn.openenergymonitor.org/pv-diversion/introduction/choosing-an-energy-diverter.md) very useful in setting out the landscape rather well. I found very little actual academic research on the topic but there are more commercial devices out there than I had suspected although few give away much about the technology they use.

[Radian]

I find it quite fascinating. The project is a nice intersection between the full stack of EE disciplines. I think I'm covering all the territory by having a LAN-centric design which gives me good visualisation and control and a chance to flex a bit of analogue design muscle because the DIY projects I've seen tend to just poke a resistor biased transformer output straight into an ADC input. I'm making full use of the air-gap afforded by the WiFi connection and 'riding' the mains voltage directly. The mains is also being used as a low-latency communications channel with the power switch which is, I think, quite novel.

 

 

The prototype unit in development.

 

Even though the circuit potentials are referenced to the mains neutral, I'm still using a split secondary transformer to derive separate DC voltage supplies for the analogue and digital sections. The PZEM power measurement unit is bottom right. I'm only using it to confirm my results and may remove it eventually. I'm sharing the same Current Transformer that came with it. Just to the left of the ESP32 module is an 8-pin DIL mosfet gate driver that makes an excellent 120kHz mains modulator (coupled with a 0.1uF X2 rated capacitor). Not much else other than a couple of linear regulators and a dual OP-Amp for signal conditioning.

[MikeSharp01]

Very tidy, if you don't count the 45 degree link on the veroboard😁, where would we be without veroboard I wonder? 

[Onoff]

23 minutes ago, MikeSharp01 said:

where would we be without veroboard I wonder? 

 

Point to point!

 

[Radian]

Point to point is how I've built the Triac load driver in the absence of a PCB:

 

 

The 120kHz signaling is up and running now. I've tested it driving the Triac opto coupler (MOC3041) direct, i.e. the measurement unit puts a 1ms burst of 120kHz just before the zero crossings of each cycle it wants 'burnt off' and the above unit can plug in anywhere around the house to switch those cycles into a suitable resistive load.

 

 

The 120kHz signal is represented by AC1 in the above circuit. It's actually the secondary of a pulse transformer tuned to the signal, with the primary coupled to the Live mains via an X2 capacitor and fusible resistor (not shown). After a bit of gain the signal goes through a high Q band pass filter followed by an envelope detector. So this circuit along with the Triac could form a self contained, remote controlled load dumper. I could easily and cheaply reproduce it and dot them around anywhere with a suitable load.

 

However, it's not gated by any other enable apart from the signalling bursts. It hasn't yet fired off on any noise and the quiescent output from the signal conditioning stages looks very quiet - but I think I'll add an ESP32 to the unit to use MQTT just to verify that load dumping is supposed to be taking place. Pity really as it's so simple otherwise.

Edited May 31, 2022 by Radian

[markocosic]

Elegant! Like.

 

Could also be the basis of a very flexible "divert" product subject to flicker/harmonics regs.

 

Ethernet-over-Power would allow additional security, without reliance on WiFi, albeit at significant additional expense.

 

Generic RF-over-Power might work for same in a lower cost deployment.

 

Have you ever heard of the old Cyclo systems at the Barbican etc? Drop a brief dead short across the 415V side of the 11 kV<>415V transformer near a zero crossing to switch the underfloor heating / immersions etc. Somebody at the LEB had fun with about 100 MW of demand response in the 70s with this approach. 🙂

[Radian]

15 hours ago, markocosic said:

Could also be the basis of a very flexible "divert" product subject to flicker/harmonics regs

 

Thanks for your comments. What I've cobbled together meets my needs quite well and the burst firing keeps harmonics to an absolute minimum. As you know, the downside of burst firing is flicker but just how much of an issue is when the diversion is only going to take place on bright sunny days. Yes it may not meet acceptable volt-drop limits but I'm not looking for a commercially viable solution here. Out of curiosity I might nip over to next door's house with a scope to see how far both the X10 and volt-drop reach. Luckily they have all three phases unlike my singleton.

 

I'm now adding an ESP32 to the load switch so I can monitor the actual energy being diverted and qualify the mains signalling. It will also provide a fall-back if the mains signalling isn't working: for display purposes I get an update over MQTT once a second for the number of Joules in the bucket so if it looks like it's about to overflow it can fire-off the Triac and send me an alert that it's in belt & braces mode. I did try using MQTT at 10 samples per second but it bogged down the whole process. I'm using AsyncMQTT but haven't tried putting the sampling into an independent task.

 

Having some intelligence in there also means I can eventually prioritise any number of diverted loads from a schedule. Any batteries would probably be a higher priority for example.

 

Not heard of the Cyclo system, most intruiging! I'm always amazed that you can poke a few +/- volts onto the mains and not have it totally absorbed by the scores of loads around a house. I wouldn't want to try putting a brief dead short on the supply though!

[Radian]

All finished now.

 

 

 

 

And an animated MQTT dash info panel showing the Joule bucket filling and emptying:

 

In the final configuration the Black box with the CT clamp goes next to the Electricity supply meter and has no external indicators or controls. Everything is done via MQTT over WiFi (where the power measurements are broadcast) and via real-time X10 style signalling to tag 'cycles to divert' on the mains wiring around the house.

 

The Heatsink box goes in the airing cupboard next to the hot water cylinder and has a single LED showing green (standby) yellow (Joule bucket filling) red (Joule bucket emptying). All told very satisfying and at a materials cost of under £30 (many parts already lying around) quite a saving on buying an eddi/harvi combo. But I don't want to add up all the hours spent building it!

 

[Radian]

Reviewing the design of the 120kHz modulating signal that I use to tag mains cycles for diversion I've come to realise that there's a perfectly harmless way to couple the microcontroller to the AC without a a transformer. The inspiration is the typical front-end of a SMPSU like the one I have providing 5V for the circuit in the Triac box shown previously:

 

 

The two blue capacitors are Class Y EMI supression capacitors wired between L & N to Earth. They are low capacitance (2.2nF) high voltage withstand (4kV) and being class Y are designed to fail open-circuit. Hence this PSU was originally in a small DVD player, with no Earth wire, and that screw coupled the common end of the capacitors to the unearthed metal case. All pretty standard:

 

 

So at 50Hz the reactance of the 2.2nF capacitors (Cy) is approx. 1.5 mega Ohm therefore a tiny 160uA leakage current normally flows to earth. Higher frequency noise especially RF noise, however, sees a lot lower impedance shunting it to ground. But at 120kHz the reactance is around 600 Ohm.

 

Intercepting the path of the capacitor to ground with an inductor (to bypass the 50Hz) in parallel with a circuit to detect the 120kHz signal results in an extremely simple receiver:

 

 

Likewise the transmitter AC couples via a Y class capacitor. The microcontroller circuit and everything on the "earthy" side of the capacitors remains referenced to earth (along with any attached PCs or test gear) and so long as adequate separation (creepage) is observed where the 240VAC joins the fun - handling the low voltage side is no more hazardous than handling a typical CD/DVD player in a metal case.

 

The back-to-back Zeners are there to clamp transients caused by mechanical switches that can interrupt the mains and produce short lived spikes that get through the capacitors. This limits the voltage to keep well within the breakdown maximums for the surrounding components.

 

I just thought I'd write this up here for anyone that might be interested.

[SteamyTea]

9 hours ago, Radian said:

just thought I'd write this up here for anyone that might be interested.

Interested, but don't understand the details much.

[MikeSharp01]

9 hours ago, Radian said:

therefore a tiny 160uA leakage current normally flows to earth

So if I have enough DVD player like loads connected I will never be able to set my RCCB / RCBO.

[MikeSharp01]

9 hours ago, Radian said:

I just thought I'd write this up here for anyone that might be interested.

Yes - really interesting thread this one. Did the heat sink come from the bits box as it seems a tad overscale or is it just me.

[Radian]

41 minutes ago, SteamyTea said:

Interested, but don't understand the details much.

It's really as simple as thinking about the analogous mechanical components Inductors and Capacitors

The Capacitor is offering very little resistance to the 120kHz signal but blocks the 50Hz. The inductor does the opposite in the same way mass and springs react.

 

Really the point I'm making here is that there's a lot of fuss (mostly rightly so) made about connecting low voltage and high voltage systems but capacitive coupling with appropriate spec. components can be considered as safe as transformer isolation. My instinctive caution is satisfied by my observation that Class Y capacitors are designed to allow weak coupling between mains live and touchable conductive surfaces. The bonus is that they provide strong coupling for high frequency signalling.

 

48 minutes ago, MikeSharp01 said:

So if I have enough DVD player like loads connected I will never be able to set my RCCB / RCBO.

Around 180 DVD players = 30mA leakage.

 

47 minutes ago, MikeSharp01 said:

Yes - really interesting thread this one. Did the heat sink come from the bits box as it seems a tad overscale or is it just me.

Yes I know WAY OTT! The thing has been sat around looking for an application for years and when I tried it on the lid of the box it was exactly the same length so I concluded it was 'meant to be'. Even With a forward voltage drop of 1V @ 16A it should barely register the 16Watt dissipation.

[SteamyTea]

35 minutes ago, Radian said:

1 hour ago, SteamyTea said:

Interested, but don't understand the details much.

It's really as simple as thinking about the analogous mechanical components Inductors and Capacitors

Is this how cheap 230V LEDs are powered?

[Radian]

2 hours ago, SteamyTea said:

Is this how cheap 230V LEDs are powered?

Yes, the reactive impedance of a chosen capacitor at 50Hz is what limits the current in the LEDs - exactly analogous to the choosing of a current limiting resistor for DC operation.

 

So for a made-up example of a string of 30 white LEDs in series (many individual LEDs integrate three or more in series on a single chip so you might only count 10 on the face of a lamp) might have a forward voltage requirement of 100V, leaving 140V to be lost across the capacitor. The AC passed by the current limiting capacitor is full wave rectified and smoother by an electrolytic capacitor to make DC for feeding the LEDs without flicker.

 

In this example, to drop 140V@0.02A (the typical forward current for a LED) requires a capacitive reactance (Xc) of 7000 Ohms at 50Hz so withXc = 1 / (2 * π * f * C) C = 0.5uF

 

If you were to try that with a resistor it would dissipate 0.02²x7000 = 2.8Watts whereas the capacitor dissipates nothing. A couple of additional resistors do usually make their way into the design - a high value (eg 1 mega Ohm) across the capacitor to bleed off any remaining charge when the lamp is removed, and a low-ohmic series resistor (eg a few hundred Ohms) to limit the current from sudden spikes that could pass through the capacitor - especially at switch-on (if your switch closure coincides with the peak of the AC sinewave). Here's one I just designed specially for you (C1 is the current limiter described above):

 

 

Capacitive droppers are pretty crude!

Edited June 15, 2022 by Radian

[SteamyTea]

18 minutes ago, Radian said:

Here's one I just designed specially for you

What software do you use to sketch things up in?

 

We did, is it LRC circuits, this in college. Actually in a mathematics lesson, got the basic concept i.e. you can reduce voltage and power with less waste heat, but did not really understand what was going on at 'the mechanical level.

[Radian]

4 minutes ago, SteamyTea said:

What software do you use to sketch things up in?

LTSpice - wonderful bit of free software from Linear Technologies. It simulates circuits to save you blowing things up for real!

[SteamyTea]

6 minutes ago, Radian said:

LTSpice - wonderful bit of free software from Linear Technologies

I like a bit of spice. Shall have a look when I get out of St. Ives. 

I may be some times. For some reason when people get here, their cars get 2 feet wider in their minds.

Edited June 15, 2022 by SteamyTea

[MikeSharp01]

36 minutes ago, SteamyTea said:

For some reason when people get here, their cars get 2 feet wider in their minds.

Its a scale thing - Kernow is such a small place, and so crowded at this time of year, that cars seem bigger in proportion to the wide open spaces of Balham - itself gateway to the south.... 

[SteamyTea]

40 minutes ago, MikeSharp01 said:

Kernow is such a small place,

One of the biggest counties in the land. Population density is a third of Buckinghamshire.

 

Think it is more to do with mean age this week. Newly weds, just breds and nearly dead.

Edited June 15, 2022 by SteamyTea

[MikeSharp01]

33 minutes ago, SteamyTea said:

Kernow is such a small place,

Good to know that irony sometimes does not hit the mark.

[Radian]

On 17/05/2022 at 08:18, ProDave said:

Before that I did experiment with a single current clamp and comparing the phase of the current waveform to the voltage, but although it worked, some loads, I recall the washing machine in particular, really confused it.

 

 

Your comment back in May stuck in my mind for some reason.

 

I have no obvious explanation for it but recently I put the Triac dump box in our kitchen to exercise it with a 3kW load (the kettle). Yesterday I noticed some odd flickering of the activity LED showing short bursts of power being dumped when it shouldn't be. It was very obvious that it coincided with the washing machine doing a fast spin. At first I thought it might be throwing the power measurements but it wasn't. Looking at the logs it was just a massive amount of noise getting onto the AC. This was more than my analogue filtering could reject so it looked like the AC cycles had been 'tagged' for consumption.

 

It should also have been rejected by the fact that the MQTT messaging didn't show any export going out into the Joule bucket but this revealed a silly bug I'd put into the code... the whole point of all this testing! I also realized I had way too much gain in the analogue path so dropped the filter Q from 20 to 4 and gain down to one. Still get a ~5V pulse corresponding to the 120kHz signal coming from across the house. I hadn't paid much attention to the fact it was clipping at this level before so there were unhelpful amounts of gain in the initial analogue design.

 

 

Anyhow, it made me recall your early troubles. It's made me wonder if the kettle might be a useful destination for excess generation at certain times of the day...

[SteamyTea]

12 minutes ago, Radian said:

made me wonder if the kettle might be a useful destination for excess generation at certain times of the day.

If your usual storage is full  i.e. thermal store and lithium batteries, would it not be better to save it as separated ions and electrons in any old battery. The efficiency becomes secondary to the marginal increase in capacity.

Or, if you are a 'prepper' and like to grow vegetables, use it to run so grow lights and irrigation, store those joules as carbs.

Edited June 22, 2022 by SteamyTea

[Radian]

 4.18 J/g°C x 1000g x 80°C means it takes 334400 Joules or 0.1kWh to boil the kettle. Probably do  5 boils between 11AM and 4PM so 15p to save. No, it makes no great sense as an isolated exercise but I'm building a list of places where I could prioritise the diversion of excess energy so I'm giving everything a chance.

[SteamyTea]

8 minutes ago, Radian said:

 4.18 J/g°C x 1000g x 80°C means it takes 334400 Joules or 0.1kWh to boil the kettle.

Mug of tea takes 0.3 kg of water.

So you are boiling enough to make 16 mugs.  The remainder can be used to wash the mugs and spoon up in.