For example: L1 has a 48/2000, L2 has a 48/3000, and L3 has a 48/5000. These are not configured as a 3-phase cluster. They are independent of each other so that if one unit fails, the other two continue to operate normally. My goal is redundancy, not a synchronized 3-phase grid. An AI confirmed this is possible—can you verify the practical limitations? (Actually I will buy 3*48/3000 now, but maybe later I will replace one or two for bigger one so single phase-setup is my target). ESS mode only. No AC-Out / UPS functionality.
Here’s example made by AI:
Here is Node-RED logic that makes your three “independent” MultiPlus-II 3000 units work in perfect harmony. Because the standard Venus OS ESS can only control one master unit at a time, Node-RED acts as the “conductor” here.
-
Preparations
Install Venus OS Large (the version that includes Node-RED).
Connect each MultiPlus with its own MK3-USB adapter to the Raspberry Pi.
Make sure every MultiPlus has the ESS assistant installed (via VEConfigure).
Set the ESS mode on each MultiPlus to “External Control”.
-
Core logic (Function Node)
This code reads the grid meter value (Grid Power) and divides it equally among the three inverters.
If the meter shows +1500 W (exporting to grid), each inverter gets the command to charge the battery at 500 W.
Put this in a “Function” node:
javascript
// Get grid power from meter
// Positive = export (charge battery), Negative = import (discharge)
let gridPower = msg.payload;
// Number of inverters
const n = 3;
// Calculate setpoint per device
// Round to whole watts – Victron prefers integers
let setpoint = Math.round(gridPower / n);
// Safety limits (3000VA MultiPlus can handle ~2400 W continuous)
const maxPower = 2400;
const minPower = -2400;
if (setpoint > maxPower) setpoint = maxPower;
if (setpoint < minPower) setpoint = minPower;
// Return three separate messages – one for each output
return [
{ payload: setpoint }, // Inverter L1
{ payload: setpoint }, // Inverter L2
{ payload: setpoint } // Inverter L3
];
- Node-RED flow structure
Build the visual flow like this:
Input Node: Victron Energy → Grid Meter → L1+L2+L3 PowerFunction Node: the code above (configured to output 3 messages)Output Nodes (3 pcs):Victron Energy → MultiPlus (Instance 1) → ESS Control → AcPowerSetpoint
Victron Energy → MultiPlus (Instance 2) → ESS Control → AcPowerSetpoint
Victron Energy → MultiPlus (Instance 3) → ESS Control → AcPowerSetpointTip: The “Instance ID” of each inverter (usually 257, 258, etc.) can be found in the Venus OS device list. Each has its own ID because they are separate USB devices.
- Why this is better than the “official” suggestion
Redundancy: If one USB cable comes loose or one inverter fails, the other two continue reading the meter and adjusting their share.
Symmetry: You stay at net-zero for the whole house, even if the phases are unbalanced.Scalability: If you buy a fourth inverter, just change n = 4 in the code.
Your scenario does not entirely make sense as to why you would then need different sized inverters.
If you are wanting to use ESS then each inverter needs a GX that is how the Victron are designed to work. Ideally then you would also have separate battery banks for full redundancy as often it is the battery bank that causes a system shutdown.
I won’t comment on the code. That then re introduces a single point where the syeten can fail and that does not meet with redundancy requirement.
Use the CT on each inverter system and keep them independent and at net zero.
As an FYI you can have an asymmetric installation if you aren’t using ESS assistant. But they do still need to be the same HW revision and model.
So you could have 3 3kVa in three phase and one phase with a unit in parallel for the phase with the higher load.
I think there is a misunderstanding regarding my architecture. I am not looking to build a synchronized, multi-phase cluster using the internal VE.Bus communication cables (RJ45).
In my setup, each inverter will operate as a completely independent single-phase node. They will not be connected to each other logically or physically on the DC or communication side.
Because of this:
Hardware revisions do not need to match. I can have an older MultiPlus on L1 and a brand-new model on L2.
Models can be different. I can run a 3kVA on L1 and a 5kVA on L3 without any compatibility issues.
The ‘Leader’ is external. The logic is handled by an external GX device (Raspberry Pi) and Node-RED, which treats each inverter as a separate asset.
This approach completely bypasses the limitations of the ‘ESS Assistant’ and the requirement for identical hardware, while providing the redundancy I need.
Your logic is like saying that if I want to own three cars, they all must be the exact same model, year, and engine size.
No misunderstanding. You mentioned it as part of your design.
For ess assistant to be controlled by a GX it has to be either in one system or have its own GX. It is mentioned in the ESS manual.
Ok, thanks! I’ll start with a standard setup and upgrade to Node-RED later when needed. I’d appreciate comments from any Node-RED users who have experience with this.
- Daisy-chained installation (Standard 3-phase setup)AC side:
The inverters are connected to their own phases (L1, L2, L3). In the main electrical panel, each has its own fuse. This part remains unchanged.Data side:
The inverters are daisy-chained together with RJ45 cables (VE.Bus). Only one MK3-USB adapter is connected to the Raspberry Pi.Limitation:
At this stage, all devices must be identical (same model and hardware).2. Switching to independent control (Node-RED)When you replace one inverter with a larger one:AC side:
Remains the same (unless the new inverter’s power requires thicker cables or larger fuses in the panel).Data side (this is what changes):
Remove the RJ45 cables between the inverters. They must no longer communicate with each other.
Add a separate MK3-USB adapter for each inverter.
Now the Raspberry Pi sees them as three separate devices.What does “changing the cables” actually mean in practice?
If you build the electrical panel and inverters now, the electrician does not need to change the fixed wiring later. You only need to:
-
On the DC side: Disconnect the shared DC busbar (if you want separate batteries for each) and run new cables from the batteries.
-
On the data side: Unplug the existing RJ45 cables and plug in the MK3-USB adapters instead.
Architecture comparison
| Feature |
Daisy-chained (ESS Assistant) |
Independent (Node-RED) |
| Devices |
Must be identical (model & hardware) |
Any (e.g. 3kVA + 5kVA) |
| Fault tolerance |
One fails → all go down |
One fails → others continue |
| Control |
Automatic (Victron) |
Custom (Node-RED) |
| Battery setup |
Shared DC busbar |
Can be shared or separate |
Scroll down to Connecting multiple VE.Bus systems to a single Cerbo GX (yes i know youhare not using the cerbo but the software limitation will apply since it is the same OS)
Option 1: Use an MK3-USB interface
This method allows multiple systems to be visible, but with limited functionality:
Only the system connected to the built-in VE.Bus ports contributes data to the Overview pages.
All connected systems appear in the Device List and are included in VRM energy statistics.
DVCC and ESS control applies only to the system connected directly to the built-in VE.Bus ports.
Additional Systems connected via MK3-USB do not support DVCC control and will follow their own internal configuration for charging/discharging.
There is an option (Settings → System Setup → Charge Control) to enable BMS control of a secondary inverter/charger connected to a GX device via an MK3-USB interface. When enabled, the secondary device will follow the CVL and DCL values provided by the CAN-connected BMS.
Generator start/stop logic applies only to the directly connected system.
For ESS systems, only the system on the built-in VE.Bus ports participates in ESS mechanisms. Others are visible in the Device List only.
Looking at your idea, i think you will get some weird oscillations sharing one grid meter but you seem pretty determined to implement this. I will add the usual caveat - remember this is a very unsupported set up.
I’ll start with a standard 3-phase cluster for now to get the system up and running with my current LiFePO4 bank.
However, in a couple of years, I plan to switch to three independent single-phase setups. I want each MultiPlus-II to have its own dedicated battery bank to accommodate different, cheaper chemistries (like Sodium-ion) on separate phases. Since 3-phase synchronization doesn’t allow for independent DC buses and mixed charge curves, the split-phase architecture is the only way to stay modular.
I’ll handle the orchestration via Node-RED when the time comes to break the cluster. Just wanted to confirm the hardware path is viable.
I hope that Victron should officially support this setup in the future.
The issue is actually the battery control…
Three different batteries with different chemistry and DVCC. And one cerbo.