Solar PV + battery - do I need a diverter?

ben.jones12 · 2024-09-16T20:32:54.462Z

Hey David,

Unfortunately the house is passive, so has an air tight membrane around the outside. The garage is outside of the thermal envelope so I have run various CAT6 and power cables from the garage to inside the house, and then sealed them all up where they pass thru that membrane.

In other words, it is not possible to run additional cables between the two locations :(.

I have been testing the system over the last few days, without the PV router enabled. I have just added a crude form of diversion using a NodeRED flow which checks the Powerwall SOC and solar generation every minute. If > 80% SOC and > 3kW of solar being generated, turn on the HWC element, off otherwise.

The HWC element draws 3kW so assuming solar export is just 3kW it will cover the HWC and the house load will be covered by the battery. If the battery drops below 80% the HWC is turned off, and any excess solar is used to charge the battery again.

This system appears to be working quite nicely. The battery is used as a buffer and ensures we never pull anything from the grid.

I am still using the Arduino with the SSR to control the HWC load, but it is not being driven by any fast PV diversion logic.

This means I can remove the PV router from the garage and recommission it inside as a standard 4x channel energy monitor (EmonTX) for monitoring 4x additional circuits inside the house.

Personally I am pretty happy with this setup now. It has fewer “moving parts”, but still ensures I soak up any exceess PV without ever importing from the grid. It is very easy to tweak the battery/solar thresholds in NodeRED if I need to.

I came here to ask you clever chaps if anyone had tried this kind of setup, using a battery instead of the PV router. It seemed like a good idea but I am no expert and wanted to see if anyone else had tried it or if there was anything I was missing!

I really do appreciate your feedback and insights!

dmajwool · 2024-09-16T22:42:56.201Z

ben.jones12:

Unfortunately the house is passive,

I would never consider that a misfortune!

ben.jones12:

The PV router sends MQTT messages over my home LAN network.

ben.jones12:

I have run various CAT6 and power cables from the garage to inside the house

Does one of these cat6 cables run from the Mk2 into the house? If so, could we move things around and use 1 pair of that cat6 to extend the Mk2’s grid clamp so that the Mk2 can go somewhere on the HW tank’s spur.
If we then temporarily disable all your 80% SoC smarts and load a basic sketch into the Mk2, we could connect the 240v inputs and outputs of the Mk2 in parallel with your switch.
I reckon that, irrespective of the battery SoC, whenever there is grid feed-in the Mk2 will start diverting into the HW tank.
If your other switch is open circuit, the energy sent to the tank will be capped by the Mk2 such that grid=0 watts.
If you close your other switch, the HW element can draw whatever power it likes, uncapped by the Mk2.

ben.jones12:

if there was anything I was missing!

Do you know if your battery “leaks” into the grid? My battery is set such that it attempts to maintain the grid feed at 0 watts. But what I find is that this is not 100% achieved and I typically import about 0.8kWh/day on a sunny day when I have plenty of solar and battery. This isn’t a big deal but it is irksome.
I find that the battery has a response time of a second or so when a load changes abruptly. Without a diverter this exhibits as frequent small imports and exports.
I find the Mk2PV router is excellent at trapping those exports and routing them into the Hot Water tank.
It’d be nice to think that the above wiring topology also addresses this in your situation as well as it does in mine.

ben.jones12 · 2024-09-16T23:01:19.796Z

dmajwool:

I would never consider that a misfortune!

Indeed, after re-reading that I realise it wasn’t the meaning I intended!

dmajwool:

Do you know if your battery “leaks” into the grid?

Not that I have noticed. I have found the Powerwall is pretty responsive to shifting loads. My energy supplier has a pretty nice visualisation of my usage (black is 0kWh);

image738×565 25 KB

My off-peak rates are between 9pm - 7am so you can see the battery is doing a pretty good job over the last month (which is late winter/early spring here).

dmajwool:

Does one of these cat6 cables run from the Mk2 into the house

I do potentially have a spare CAT6 cable I could re-purpose for this. But I am not sure I see how it would help. The HWC load would still be “inside” or “behind” the Powerwall Gateway, so it is going to see that load when it is on, and switch to battery if there isn’t enough solar to cover it.

I think I need that HWC spur to be before the Gateway in order for the PV router to work properly… I think…

Moving the PV router next to the HWC doesn’t change anything I don’t think. Because I am using MQTT/LAN to control the HWC load (via a SSR) they are “virtually” next to each other. The PV router sends divert on/off signals to the SSR over MQTT, but it is in effect directly controlling the HWC load and can switch it fast enough to balance things out.

If I could re-route the HWC spur into the garage and wire it into the switchboard “before” the Gateway, i.e. between the grid/meter and the Gateway, I think it would then make a difference and allow the PV router to accurately soak up any excess being exported to the grid.

dmajwool · 2024-09-17T07:33:39.690Z

Hi Ben,
Lets strip this back a bit. Take another look at Mk2PVRouter - Introduction
In the block diagram
39cab288cf56fee4ffff8041d4355564
Consider that the “consumer unit” block is your Powerwall Gateway and that your battery is 100% full and the sun is shining.
Also consider that the Mk2router is inside your house adjacent to your existing smart switch and wired in parallel to it (so that when closed the smart switch bypasses the Mk2router).

In this scenario, your PV inverter will be sending power to the Gateway which will then be flowing out to the grid.

My Mk2router doesn’t have any nodered, mqtt or any such stuff. It is entirely stand-alone. It has a grid c.t., an Arduino sketch and a solid state power gate that opens and closes, conditional on the grid c.t.

The 240 volt supply wiring route into the Mk2 does not matter. Anywhere inside your gateway is fine. All that matters is that the c.t. registers some grid export and then the Mk2 increases the duty cycle of its gate (delivering energy to the load) and the grid c.t. sees correspondingly less export until a balance is found and your grid export reduces towards zero. (unless the water is already hot or if you have more than 3kW excess etc etc)

When the c.t. detects import from the grid (ie when the battery is charging from the grid), the Mk2 gate stays open circuit.

The battery is not connected directly to the HW heater so there is no reason why the battery will runaway and supply maximum power to the HW heater. The battery is connected to two switches in parallel - the Mk2 and your smart switch. The Mk2 switch will only activate when energy is flowing out to the grid, and your parallel switch will do whatever some other device (your finger?) instructs. So you can program some timer or SoC dependent water heating into this smart switch.

When wired up like this (as I have), the battery has first call on the PV energy. It will charge the battery before diverting excess to water heating. This is not entirely efficient because of battery round-trip losses. In my battery there is a register to limit the charge amps. When the season allows, I set this to a lower figure in order to force some export which will then be picked up by the Mk2. We find this convenient because the water is heated earlier in the day.

ben.jones12 · 2024-09-17T20:48:47.680Z

Hi David,

I am thoroughly enjoying this discussion by the way! Thanks for taking the time!!

I think there is still some misunderstanding about my setup. I wish I was better at drawing so I could map it out more clearly.

My PV router is physically located in the garage, with a CT clamp around the grid feed, between the supply meter and the gateway. So it is correctly measuring the grid import/export, as in the diagram above.

The PV router is also directly controlling the power to the HWC, albeit via virtual connection over MQTT/LAN. The net effect however, is that as the PV router detects export to the grid, it will switch the HWC element on/off at a rate that brings the export back to zero. This all works and I have been using this system for years in my previous house, with no gateway/battery.

My PV router is the same as yours, standalone. It has custom firmware to send the load on/off commands over MQTT rather than switching a relay directly via a GPIO pin, but again, the net effect is the same. It is directly controlling the delivery of power to the diverter load, my HWC element.

It also has the ability for me to disable the diverter, via an incoming MQTT message. So I can tell the PV diverter to stop trying to divert if I wish. This doesn’t do anything other than prevent the divert on/off signals from being sent.

So aside from that, the PV router is very much like the diagram above, identical in fact.

So the PV router will turn on the HWC when it detects export to the grid - i.e. when the battery is charged and there is solar being generated, but also when there is a rapid change in load in the house and the battery might take a second to react meaning there is a small amount of export as it re-balances itself.

When this happens the HWC element turns on and starts drawing 3kW. If the solar generation is only 2kW and the house is using 1kW then there is a 2kW shortfall. At this point, assuming we are in the peak rate tariff (i.e. day rate), the gateway will try to avoid any import and start discharging the battery to cover that shortfall.

The PV router CT clamp sees the grid feed return to zero, since the battery has now taken over, but it doesn’t see it start importing, therefore it never sends the OFF signal to the HWC load. It is happy that the diverter load is perfectly balancing out the grid feed to zero.

So, because the diverter load sits behind the gateway, once the PV diverter is triggered by any kind of grid export, that load will remain triggered until the battery is depleted enough for the gateway to start pulling from the grid (or if there is a rapid load change in the house which causes a short spike of import), which the PV diverter will see and then shutoff the load.

As per Roberts original diagram above, if the diverter load was “before” the gateway, then the gateway would never “see” that load and the battery would never be used. Thus the diverter would be able to happily switch the load as required to ensure anything coming “out” of the gateway, and destined for the grid, can be soaked up before being exported.

This whole process has been very useful for me, writing these sorts of things down always helps get a clear picture in my mind of what is going on!!

Robert.Wall · 2024-09-17T22:03:24.243Z

Going back to what I wrote in Post No.18, the problem is that Gateway was never designed to be part of a bigger system – one incorporating the Mk2 PV Router. Neither was the Mk2. What you have are two control systems each forming what we call a ‘closed loop’ - meaning the system monitors the result of its actions and modifies that action accordingly. What you (@ben.jones12) have --and you’ve realised this – is two control loops that are intertwined.

My thinking is turning to asking what information do you have or can find out by inspection about the internals of the Gateway? Is it possible to reroute cables and locate the Mk2 c.t. inside, maybe adding or subtracting currents by passing wires in different directions through the c.t. so that it sees a current that will give it the information it needs. Can you do what @dmajwool did in 2019?

You also mention that the Mk2 responds much quicker than the Gateway (as it’s designed to do), so would holding off its response in order to give the Gateway time to pick up or dump its load (to or from the battery) solve the problem – but at the cost of a few seconds of imported or exported power?

dmajwool · 2024-09-17T22:16:40.868Z

Thank you Robert.

Ben’s latest post reminded me that you helped me with some wiring tweaks to avoid my battery getting into the same kind of loop where it supplies more and more power to the dump load.

Our solution was to give the battery’s grid c.t. a sniff of what the dump load was receiving, as discussed in the 2019 thread.

In Ben’s situation this may be more difficult to achieve because of the separation between the dump load and the grid monitoring point. But in addition to the 2019 post you quote, there is also a post in 2023 where we discuss using multiple physically distant c.t.s to aggregate a value. And that might be a good approach in Ben’s situation. aggregating multiple c.t.s

Robert.Wall · 2024-09-17T22:41:16.715Z

dmajwool:

Our solution was to give the battery’s grid c.t. a sniff of what the dump load was receiving, as discussed in the 2019 thread

No we didn’t - we removed the current the dump load was using from the total the grid c.t was seeing by running the cable carrying that current backwards through the c.t. – subtracting it from the total.

But we know @ben.jones12 can’t do this because of the topology, so we’ve got to work out an alternative - maybe add or subtract something from the Mk2 c.t?

dmajwool · 2024-09-17T22:49:00.716Z

dmajwool:

a sniff of what the dump load was receiving

Sorry, you are quite right. Just my lazy writing.

dmajwool · 2024-09-17T22:55:29.164Z

Robert.Wall:

But we know @ben.jones12 can’t do this because of the topology

Perhaps it is possible (if the Tesla Gateway uses a c.t.) to combine that c.t. with another around the dump load spur as you suggest here

Aggregate reading from multiple c.t's?

If you are using genuine current-output c.t’s, then connecting them in parallel results in the currents adding (or of course subtracting if you reverse one of them - either with the direction of the primary (the cable) or with the secondary connections). If they are the voltage output type, then wiring in series will have the same effect. Both c.t’s must have the same sensitivity, or you’ll need to be inventive and you might need an additional burden for one of them, the aim being to make them appear to have the same sensitivity at seen at the input to your inverter.

We know that Ben has some twisted pairs in CAT6 cable available to extend c.t.s between garage and house.

ben.jones12 · 2024-09-17T22:58:47.435Z

I keep coming back to my original question - with a battery in the equation, that can act as a buffer for an extra PV (by charging) or extra load (by discharging), is there much need for a diverter?

I can see that with the right wiring you could definitely still use a PV router, but even in this scenario is there much benefit? Assuming the battery can regulate itself quick enough to minimise import/export.

Robert.Wall · 2024-09-17T23:07:10.994Z

ben.jones12:

is there much need for a diverter?

You thought there was before the Gateway appeared, what you’re really asking is, what happens if I don’t have it? I think this turns into “Is my battery big enough to absorb enough of the P.V. so that it will support all the normal stuff in the house and the immersion heater if it’s on a thermostat / time clock or whatever?”

ben.jones12 · 2024-09-17T23:19:22.891Z

Robert.Wall:

Is my battery big enough to absorb enough of the P.V. so that it will support all the normal stuff in the house and the immersion heater if it’s on a thermostat / time clock or whatever?

Agreed, that is what is boils down to. And since turning off my PV diverter I think the answer is “Yes” in my case. Particularly once I add some basic automation rules over the top, where I only turn on the HWC load if the battery > 80% and solar generation > 3kW.

I really do appreciate your expert opinions and insights @Robert.Wall and @dmajwool. I think I am happy I have considered all the options and am not missing anything.

I hope this thread is useful to someone else in my position with similar questions

dmajwool · 2024-09-17T23:23:49.625Z

ben.jones12:

is there much need for a diverter?

If you have more PV kWp than the battery is able to absorb.
If you are concerned about round-trip efficiency of the battery (you can measure this - with AC coupled PV, mine has significant losses)
If you are concerned about the total lifespan (joules of throughput) of the battery cells.
If/when the battery goes offline.

I asked myself the same question when I added a battery to the system, but I’m pleased I kept it and the Mk2router remains a useful part of my total system.

ben.jones12 · 2024-09-17T23:38:11.282Z

All valid points, but in my situation it just doesn’t work effectively unfortunately, thanks to my wiring topology.

I wish I had cottoned onto this nuance earlier so I could have run some extra flex cables between the garage/house, so I could isolate the diverter dump loads from the gateway.

But that would introduce another issue, if my grid connection went down and the gateway switched to battery backup, I would not have power for my HWC.

Although that could be solved thanks to having 2x elements in my HWC (middle and bottom). The middle one could be “inside” the gateway and not part of the diverter load, the lower one would be the diverter load and not part of battery backup.

I will know for the next time I build a new house

dmajwool · 2024-09-17T23:57:58.006Z

Hi Ben,

ben.jones12:

but in my situation it just doesn’t work effectively unfortunately, thanks to my wiring topology.

This is only a discussion. Our work is done if you’ve appreciated that your power wiring topology is OK as it is, and you could achieve a different diverter result by running the c.t. clamps in a novel way using cores of your cat6 cabling running between garage and house.

Cheers, David.

ben.jones12 · 2024-09-18T00:14:09.865Z

So in order for this to work in my situation, I would need a CT clamp around the HWC element spur, which is fed back to the PV router over a spare CAT6 cable. That remote CT would then need to be plugged into the PV router and the firmware customised to subtract that reading from the grid CT readings.

Do I have that correct?

dmajwool · 2024-09-18T11:58:25.756Z

ben.jones12:

Do I have that correct?

Not quite. The suggestion is to change the wiring of the Gateway’s c.t.

Here’s my brown grid incoming live cable. My Battery’s grid c.t. is on the left and the Mk2’s grid c.t is on the right.

The additional cable running through the battery’s c.t. is the supply to the Mk2.

20240918_111821697_iOS1920×1440 224 KB

Robert.Wall:

we removed the current the dump load was using from the total the (battery’s) grid c.t was seeing by running the cable carrying that current backwards through the c.t. – subtracting it from the total.

If I’ve got things clear in my head, the logic here is:
Your present situation:

When there is excess PV, the Gateway starts exporting.
The Mk2 starts diverting.
In the design of the Mk2 software, taking advantage of the way utility meters register electricity usage, the Mk2 makes deliberate small (less than one flash of the meter’s LED) overshoots of import and export.The energy bucket
The Gateway sees the short period of import and increases its export to balance it.
The Mk2 sees this additional export and diverts more. etc etc.feedback loop.

The solution that worked for me
As Robert said,

Robert.Wall:

we removed the current the dump load was using from the total the (battery’s) grid c.t was seeing by running the cable carrying that current backwards through the c.t. – subtracting it from the total.

The effect of this is

When the battery is exporting
The Mk2 diverts, reducing the export towards zero
The modified battery c.t. doesn’t see that the export has reduced - or turned into a short tiny import.
So the battery doesn’t react to the Mk2’s behaviour.

I don’t know if the Tesla Gateway has an accessible c.t. that can be tampered with in the same way that my battery does. But if it does, I think you could try fitting a c.t. around the immersion heater spur and extending it out to the gateway’s c.t.wiring and combining them as discussed here

Or you might think the work outweighs the benefits

ben.jones12 · 2024-09-19T22:25:07.055Z

dmajwool:

Or you might think the work outweighs the benefits

I think that is where I am landing .

But I am glad to have gone thru this process with you two, I have learnt a lot and feel like I understand the system better than before.

I certainly didn’t realise you can wire two CTs in parallel in order to net their readings against each other!

Robert.Wall · 2024-09-20T01:07:28.211Z

ben.jones12:

I certainly didn’t realise you can wire two CTs in parallel in order to net their readings against each other!

With care! They must be both genuine current-output and the same ratio. This works because they are current sources, and wired in parallel the currents add.

If they are voltage-output type, both must be the same output voltage for the same rated current, and they are wired in series because they are voltage sources, and wired in series the voltages add.

You must calculate the value of the total current and size the burden accordingly; or ensure the sum of the output voltages doesn’t exceed the input voltage.

If they face in opposite directions, the currents/voltages subtract.

Or you can have a current-output c.t. with its own burden and it becomes a voltage, which you then wire in series.