Comment by archi42 | Hacker News Reader

Comment by archi42 | Hacker News Reader

Parent Full thread

For those thinking about replicating this: Victron is really popular because it's extendable and the management interface is reasonably open, but the zoo of required devices makes the costs explode quickly.

Maybe look at Deye instead. I'm running a 12kW unit (they're available smaller) with a 48V battery. The software is in poorly translated to English, firmware updates are made by the manufacturer remote on request and it has no shadow management (meaning it might end up on a local maximum in the MPPT, not the global maximum) - but the unit cost me about 3000€ with 10yr warranty instead of the >7000€ for an equivalent zoo of Victron parts.

Of course it can do black starts from PV or battery, and act as a UPS on a secondary output. Pulling data locally can be done using a RS485 modbus interface; so I open my firewall only for firmware updates.

The 3.6kW single phase unit (SG05LP1) can be had for less than 1300€. A 5kW three phase unit is only a little bit more expensive, but I'd worry about standby consumption in low power scenarios with only a few modules (mine hogs 80W, but a three phase Victron setup does so as well). These are string inverters, so you need more modules for them to reach their start up voltage, and then some more to put them into the MPPT zone. The 1 phase starts at 125V and the MPPT tracker works with 150-425V, the 3 phase units 160V & 200-650V respectively. When I browsed the market there were other offerings from other brands with 90V start up; but be careful to get a unit that's legal to connect to the grid IF you want to have grid fall back (here in Germany, Victron and Deye are legal - at least the parts I looked at. But you will always need a electrician with a concession to register them with the grid company. Some people don't care and just setup their systems though).

Important: The voltages depend on the chosen PV module; mine are in the 40V per module ballpark, but iirc I might maybe have seen up to 60V? So for the 3.6kW unit you'd want at least 3 or 4 modules.

Victron is good stuff and it is made to last. They're a big supplier to maritime integrators to provide power on board of ships of all shapes & sizes. I've yet to see a Victron inverter fail before the end of the lifetime of the installation it was a part of whereas most other inverters including Xantrex and ABB tend to die after a decade or so. Which isn't bad but still it can be a big hassle to replace an inverter if the form factor has changed.

Also: consider decoupling your battery charger setup from the solar portion, this will give you a lot more flexibility in terms of siting (you can then place your battery and charger where you want on your local installation). You can use HA to bring your battery online when it is most needed and even when to charge it based on the outputs of multiple fields of solar, an integrated solution really only works with a single inverter or a set of inverters of a single brand.

And if you can stay away from stuff that requires an internet connection to some manufacturer, not all upgrades are positives and your data shouldn't be going there anywhere. Also it probably isn't a good idea to hand over control of a grid connected device to another party, especially when that device has firmware that can't be audited.

Ah, a jacquesm reply, I feel honored :)

Of course you're right, the Victron are good, and they're proven. They're incredible flexible and for hackers like us they're amazing units.

But for me the Deye was 2500€ shipped (group order with the german tax break applied), with 10yr warranty and a local company acting as a liason for warranty claims. If the Victron setup would have been 3000 or maybe even 3500€ I wouldn't have looked at the Deye - but we're currently energetically renovating a house, and with the current price hikes we have to be careful not to drain our funds before we're done. So for me the choice was a smaller system (both kWp and kWh), or the Deye.

Since I can locally control the unit via RS485 (I use HA and node-red), I'll eventually do a lot of sheningans as well. I despise things that require cloud and the Deyes cloud connection is no different (and hence severed in the OpnSense). If people want to consume most of the generated power onsite good local control is a necessity. Once we have other things like the central heating and the new roof done and when I'm not busy with plaster work around the new windows anymore, maybe then I'll setup the more advanced energy managment like evcc or custom automations. Until then the 14kWh battery takes care of ensuring energy stays local (and I only make sure not to charge the PHEV battery from it by accident - the Deye happens to have a suitable output port that's activated if the battery is full; I have yet to wire that port though).

Anyway, for someone who has the liquidity, and feels they want to use the Victron ecosystem, by all means, for lots of metrics it's better than the Deye; but for me I didn't see that the Victron would save me more money than the initial invest. Also, the single Deye happens to fit my requirements pretty well (it's a bit more powerful than necessary, but the extra costs over their smaller units were neglibible).

It's all trade-offs. I have a pretty weird setup with lots of gear that gets switched in and out according to power availability, and inverters in different spots (two in the garage and one in the attic). I still haven't installed a battery because we - for the moment - have 100% net metering here so it would be a waste of money. TCO is one way to look at things, installation flexibility another and finally of course initial outlay: better to buy the setup you can afford than to drool over a better one that you can't.

I've yet to come across Deye in the wild so nice to have some data points. What I'm really curious about is to what degree the various setpoints are configurable. I have a Growatt here that is giving me a ton of trouble and two Solax units that have been rock solid since first installation (and without any configuration at all). All of them offline.

On the panel side here I have a bunch of AEG branded glass/glass panels (AEG has licensed their brand so these aren't actually made by AEG but they are made in Germany and look to be well constructed and with a 25 year predicted life span I think I'll be long gone before the panels will die).

I'd love to hear how you are faring with all of the stuff that you've done.

I have it running in a testing setup for about 2 weeks now. 8 modules flat on the lawn (Maysun glass/glass 410Wp), the battery connected and tied to the grid with export power limited to 0 because my meter is ancient enough to spin in reverse (which is illegal to do here).

Which setpoints do you mean? I'm not familiar with all the terms, especially when they vary by manufacturer.

If you mean the power limit the Deye should put on the DC bar (i.e. put into the battery), that works reasonably well. If it's at 100% it will still charge a few Watts, but nothing severe.

The other meaning I could come up is the target power to draw from the grid when export is disabled (as in my case, for now), or when it is feeding energy from the battery into the house when I draw more than the PV array delivers. That's 20W by default, but can be user configured. This fluctuates somewhat strongly, but overall works out correct. I believe the reason isn't the software, but the poorly routed measurement cables: It uses measuring transfomers and the cables in my setup go tightly and parallel to the power line that connects the inverter to the grid. The electrician said that's not an issue, but I guess he usually measures 50A+, in that case the noise is negligble. Around the 0A point the noise has a much bigger impact. But that's easily fixed. Eventually I'll use that option to prevent the battery from discharging when the car is charging over night (e.g. when I charge with 1200W, I plan to set the target to that value).

I'm not sure if I can configure other setpoints, e.g. the MPPT target voltage (which would be a nice trick to get shadow managment). I can define some other things like "from $x a.m. to $y p.m., keep the battery at $z%, grid charging is [not] allowed, powering on the external generator is [not] allowed". For testing the battery balancing (it has an additional Neey active balancer) I used that. And yeah, it can trigger an external 3phase generator to power up (not sure if the 1ph can do that).

As far as I could find, all options are exposed via RS485. I adjusted the sunsynk HA addon to work with the newer 3 phase inverters (sunsynk = deye rebrand; changes are already merged upstream). Only thing that I'm missing is the current limit set by the BMS, as well as other BMS data sent to the inverter (I did not search the published register map for that, yet).

What I can already say:

All in all the technical side is pretty nice; as said in a sibling comment, I could theoretically push out 12kW AC on a single phase if my local load is severly inbalanced. And the MPPTs are pretty flexible in how much power they can accept (as long as the sum is less than 15.6kW at all times).

What's horrible is the UI side. Some phrases are highly unclear, and requires knowledge of what the option actually does. E.g. "signal island mode" means "in case of grid failure, tie neutral to PE on the UPS output". In German installations that's always necessary, else RCDs will be much less effective. They would benefit greatly if they had a team of knowledgable translators.

Having to ask their support to do an update is also a joke; especially since they don't publish release notes. Also, the update power cycles the inverter, which causes the UPS output to lose power.

My 14.3kWh LFP battery has a 200A Seplos BMS. The inverter and the battery talk via CAN. In the BMS the power limit is set to x Amps. But the inverter shows a limit of (x/2) Amps. Now my battery should eventually operate a 140A continuous, but I'm not feeling to well about setting the limit to 280A and one sunny day having the two systems decide to get along and actually push 240A (the Deye's limit) over the connection+BMS sized for "only" 200A. The solution is to get another BMS, but back when we formed the respective group orders, people didn't yet know of this incompatibility. Plus we were focused on Seplos because the Victron users felt it was imporant to get a BMS that would not disconnect the negative bus.

Still, for someone looking at a small plant with just a few panels and a small battery, it's a great value preposition. Especially when paired with a different battery/BMS. With the 8 modules behind the house (shadowed in the morning and the late evening) I had 105kWh PV and 120kWh grid power over the last two weeks. That's 1k€ for the modules, 1.3k€ for the smaller 1phase 3.6kW inverter, 1.2k€ for the 3kWh Pylontech US3000C, plus some cables and AC fuses for 200€, plus a DC fuse+SPD box (<100€ if DIY) = 3.8k€ for basic stuff. What's missing is panel mounting (800€?) and, depending on local laws, an electrician to sign off on everything.

12KW on a single phase is really not what you want to do though, that works out to 50A which is likely more than your distribution panel is wired for, usually they are good up to 40A and 50A for a couple of seconds but more than that and you are inviting blowing up the utility side fuses. You will also need pretty beefy cable to be able to run that kind of current, 6 mm^2 ~30A, 10 mm^2 ~40 and 16 mm^2 ~55. Unless you want to run your cable really hot, which I would not recommend. Also, you will have to beef up the wiring in your distribution panel, the utility side fuse (which normally is 3x25 or 3x35, rarely 3x50 though it is possible but will require fatter bus bars and special breakers as well). Much better to divide the power over your phases. That way you can run 12KW on 3x2.5 mm^2.

The setpoints I am struggling with are the difference between for instance the voltage at the inverter versus the voltage at the distribution panel, when pushing a mere 5KW into the grid from the upstairs inverter that raises the voltage enough that the inverter will trip. This is not due to the usual suspects (faulty wiring, loose contacts) but simply because of the length of the feedline to the attic and the fact that it is 2.5 mm square tri phase cabling, which has a 4V voltage drop over that length and there is absolutely no way to compensate for that.

Highly frustrating. Other interesting set points would be to be able to de-rate the inverter and to be able to get it to shut down faster once the sun has gone under.

That's really good value for the money what you've got there. My setup is a bit more expensive than yours but it also makes a lot more power (as it should :) ), I've spent a lot of time on getting it all set up properly for the summer. The biggest problem remaining (and that I won't be able to fix) is that I can't angle the panels on the low roof properly because that would upset my neighbors so they are at a very shallow angle, just enough to the get the water to run off them but even then at the edge of the panels in the 'down' direction there is a little area of about 5 cm where water pools long enough for algae to grow. So I will have to clean these quite frequently. Maybe I'll figure out a way to deal with that but so far nothing that really worked well. A steeper angle would require a higher starting point or a lower ending point, neither of which are feasible.

The real pay-off is in March/April and September/October when we will save substantially on gas by heating using an air/air heatpump with a COP of about 3 to 3.5 at the temperatures at that time of the year. That should put a sizeable dent in the energy costs.

Dead thread comment: I just scored a 17 KW Solax inverter on the local ebay equivalent for $500, 4 strings, 2 MPPT, super happy :)

> consider decoupling your battery charger setup from the solar portion

Can you provide more details on this? I’m having a hard time imagining how the solar, inverter, and battery chargers are connected in relation to the house and grid.

And what is HA? Thank you.

HA -> Home Assistant

If you have a separate charger/inverter that you can control remotely using HA then you simply monitor your outgoing power and if that goes negative and your batteries aren't charged yet you instruct the charger portion to charge the batteries. Then, when the sun goes down you can order the inverter part to come to life and supply current up to the point where the outgoing current is close to zero. Like that you maximize your own usage. And by having this split up you can site your batteries in a spot where the temperature is relatively stable, but still outside of the house which is usually not where you have your solar panels.

A solar inverter should be sited relatively close to the panels to avoid cable losses and to reduce the risk of having unfused 800V DC cabling running through your house for any appreciable length. Ideally you want those cables to be just long enough to get under the roof to the spot where the inverter sits and then regular cabling to the distribution panel.

By connecting the charger/inverter combination to the distribution panel in the same manner you get a lot more freedom. And if you have multiple sets of solar panels that's really the only setup that makes sense otherwise only a limited number of panels can be used to charge the battery (the ones that are directly connected to the inverter that feeds that battery). That's why I'm not a huge fan of the so called 'hybrid' inverters. They make a lot of assumptions about the physical layout of the location that may not hold true in practice.

Here I have 50 panels divided into five fields, 6,8,8,10, and 18 panels in each of them and three inverters. This gave a lot of flexibility in where panels could be located. But if I want to charge a battery in a hybrid setup I'd have to pick which set of panels to use and then I'd still be limited in where I could locate the batteries, presumably they'd be close to whichever set of panels I would like to use to charge them with.

But by decoupling that and using the AC network as the bus I can put them pretty much anywhere that I can reach with another cable run into the distribution panel. That way all 50 panels can contribute to the charging.

An interesting extra option is to run one more cable from the battery charger/inverter to the inverter port, that allows you to power a set of critical loads from battery power. But since the grid here is very reliable I won't be using that.

hth, feel free to ask more.

I haven't heard of a setup like this before. It seems like you would be loosing a lot of efficiency having to translate the voltage to AC and back to DC to charge your batteries.

Not really, those inverters are 96-98% efficient.

Not sure what zoo of parts you’re referring to? I have two multiplus 5kva 48v, a bmv-712, a venus gx, and a pair of their biggest MPPTs - and for our pretty large install it shook out at about €4,500. Without the redundancy, it would have been more like €2,500.

Don't get me wrong, the Victron systems are nice & I envy their great traits I've praised in sibling comments. But your system doesn't compare on the power specs.

There is no exact equivalent for the Deye with Victron. The Deye can do 15.6kWp solar, split over 2 MPPT. The topology is such that one MPPT can do 15kW while the other does 0W (I did not believe this, but I've talked to the manufacturer and it's indeed correct). The inverter can output 12kW AC, or draw that much for charging the battery. If your load is not balanced well, there is a video of someone testing it with 12kW sustained output on a single phase (this is only useful for a select few maniacs though; I'm not one of these). And if the 12kW is not enough, you can put them in parallel (like with Victron).

I deleted my notes on the Victron setup, but this is something similar to what I was looking at (about Sept 2022): 3x Multiplus 48/5000 on the AC/DC end for full three phase setup (back then 1700€ each, now 1300€) - mind those can really "only" push 5000VA or 4000W at 25degC. Our roof would be 1x MPPT RS 450/200 (pretty stable at 1800€), the string in the garden 1x 450/100 (1200€ in Sept, 1000€ now). A RPi plus USB adapters and PSU and sturdy case for VenusOS (or a GX unit) would be close to 100€ extra, also add additional cables and the necessary busbars (dunno, 200€?).

For a three phase that's 7000€ today, or 8400€ in September. Of course that's just one way of building a system, and Victron is amazingly flexible - and with my 14kWh battery two multiplus would probably have been enough (caveat emptor: no idea if my grid operator would have accepted 2x5kVA or rejected it due to the inbalance).

Lots of people use a SmartShunt or Lynx, these also add to the costs of a Victron system.

Why zoo: The footprint of the Victron 3x MP and 2x MPPTs is nearly 0.9m²; add busbars and other periphals, and you're easily at 1m². Plus, each device needs some clearance due to the heat. The Deye is one unit with 0.3m² footprint. This is also nice.

Ok, you’ve got me interested - thanks for the write-up - now I just need to see if it can coexist with the victron ecosystem I already have. Probably.

Reply on news.ycombinator.com