Feature discussion: Model shared inverter capacity between DC EV charging and home battery charging (Sigenergy EVDC)

[liambresnahan]

Problem

I have a Sigenergy SigenStor with an integrated DC EV charger. I’ve noticed a consistent difference between Predbat’s predicted charging behaviour and what the hardware actually does.

Predbat can plan simultaneous EV and home battery charging, but in practice the EVDC appears to consume the inverter’s available conversion capacity, preventing the home battery from charging until the EV finishes.

This is not a whole-site grid import limitation. The constraint appears to be the shared hardware capacity within the Sigenergy system.

I’m raising this as a discussion about the prediction model rather than prescribing a particular implementation.

My setup

• Predbat v9.4.2
• Sigenergy SigenStor, nominal 10 kW inverter (Predbat reports an 11 kW inverter limit)
• Sigenergy 12.5 kW integrated DC EV charger
• Two battery modules, 18.08 kWh usable
• Battery rated charging/discharging power: 8.8 / 9.6 kW
• Sigenergy firmware V100R001C22SPC117
• EVDC Fast Charging enabled
• Battery-assisted EV charging disabled in Sigenergy
• 18 kW grid-import limit configured
• EDF overnight rate, 23:00–06:00

Predbat’s car_charging_rate is configured at 12.5 kW, corresponding to the EVDC’s rated capability.

Observed behaviour — 20/21 October

Time Predbat Actual behaviour
23:00–00:00 Hold for car Home battery held at 4%
00:00–02:10 Charging, Hold for car EV charging at ~10.5 kW; battery charging at ~0 kW
02:10 Charging EV finishes; home battery immediately starts charging
02:10–05:50 Charging Battery initially charges at ~4.8 kW, tapering later
05:50 Hold charging Battery reaches 100%

During the overlap, grid import was approximately 11.6 kW, comfortably below the configured 18 kW limit.

At 00:07, Predbat logged a control failure because it requested approximately 3.4 kW of battery charging while actual battery power was effectively zero.

The overnight schedule ultimately recovered and the battery reached its target, but the original predicted energy flows were not achievable.

What I found in the prediction code

From reviewing prediction.py and prediction_kernel.cpp:

• EV charging is represented as an additional load.
• Battery charging is separately calculated against battery and inverter limits.
• The inverter-limit calculations do not appear to account for EVDC power using the same converter capacity.

This seems appropriate for an independent AC EV charger, but potentially inaccurate for a DC charger sharing an inverter’s power electronics.

One additional complication is that the EVDC’s effective charging power depends on Sigenergy’s configuration. With battery-assisted charging disabled, I observe approximately 10.5 kW. The charger is nevertheless rated at 12.5 kW, and that higher power may be achievable with battery assistance enabled.

Simply reducing car_charging_rate therefore seems an incomplete workaround.

Questions for discussion

1. Does Predbat currently have any mechanism to represent a shared conversion-capacity constraint between a DC EV charger and the home battery?
2. Would it make sense to distinguish a charger’s rated power from its effective power under the current inverter/EVDC configuration?
3. Could the prediction model account for the available converter capacity after EVDC demand, potentially including configurable priority or power-sharing behaviour?
4. Should this be a generic topology capability rather than Sigenergy-specific logic, so independent AC chargers remain unaffected?
5. Is this best handled as an extension of the joint EV/battery scheduling work in Joint cost-aware EV and home-battery scheduling under a site import limit #5490, or as a separate constraint within the existing prediction engine?
6. Would there be value in comparing the predicted and actual battery/EV power during overlapping charge windows, and flagging persistent discrepancies?

I’m particularly interested in an approach that models the physical capabilities and constraints accurately, rather than introducing a blanket rule preventing simultaneous charging. Reduced-rate simultaneous charging may be feasible in some configurations.

Related work

• Joint cost-aware EV and home-battery scheduling under a site import limit #5490 — Joint cost-aware EV and home-battery scheduling under a site import limit
• import_limit setting requirement #2449 — Import limit setting requirement
• Discussion Limit import load #3556 — Limit import load
• Discussion Sigenergy EVDC Fixes and Workarounds (Plus V2H) #3530 — Sigenergy EVDC fixes and workarounds

Happy to provide Home Assistant telemetry, a Predbat debug capture or further details to help validate the behaviour.

[gcoan]

Does Predbat currently have any mechanism to represent a shared conversion-capacity constraint between a DC EV charger and the home battery?

No, I think you are the first person to raise this as an issue. In most cases the EV charger is separate from the inverter charger so there isn't any constraint.

Should this be a generic topology capability rather than Sigenergy-specific logic, so independent AC chargers remain unaffected?

You can achieve some of modelling these constraints through automations, throttling inverter rates using select.predbat_manual_api, e.g. when an automation detects that car and battery charging is going on at the same time, and initially its probably worth exploring that to see if it can achieve a more accurate plan for you.

But this does look like its a candidate to be raised as an enhancement request to build this into Predbat in an extensible way for other similar combined EV/inverter chargers.

It would be useful if you can get a feel for what the limitations are, e.g. is car prioritised over battery, is it just battery charging rate affected, is it a fixed limit or is it an overall total throughput that cannot be exceeded, etc. These will help shape a future enhancement design

[liambresnahan]

Thanks, that makes sense.

I’ve looked at the HA telemetry from a couple of overnight charging sessions. So far, the EVDC appears to take priority: it charges at ~10.5 kW while the home battery remains at 0 kW, despite Predbat requesting battery charging. As soon as the EV finishes, the battery starts charging. Grid import is well below its configured 18 kW limit.

My suspicion is that both share the inverter’s ~11 kW conversion capacity, but I haven’t confirmed whether Sigenergy dynamically allocates any spare capacity or simply blocks battery charging while the EVDC is active.

I’m planning a controlled test reducing the EVDC power limit to 8 kW and then 5 kW while Predbat requests battery charging. That should help establish whether it’s a shared throughput limit, EV-first prioritisation, or a hard interlock.

I’ll report back with the findings. I agree the Manual API could provide an interim workaround, but I’d prefer to establish the actual hardware behaviour before introducing additional control logic.

[gcoan]

My suspicion is that both share the inverter’s ~11 kW conversion capacity, but I haven’t confirmed whether Sigenergy dynamically allocates any spare capacity or simply blocks battery charging while the EVDC is active.

I’m planning a controlled test reducing the EVDC power limit to 8 kW and then 5 kW while Predbat requests battery charging. That should help establish whether it’s a shared throughput limit, EV-first prioritisation, or a hard interlock.

Makes sense, understand what the behaviour is

[liambresnahan]

I did some testing and when limited there is indeed charging of both the car and battery simultaneously.

The inverter appears to charge the car first and then the battery. This of course makes the planning quite different from reality so it'd be great to be able to model the shared inverter capacity as a feature request, if possible.

What would be the best next step here?