Is raising inverter_limit the right way to plan a 5 kW grid charge (plan shows charge 4.80 kW vs export 5.00 kW)?

[pietrouk]

Predbat version: v9.6.4

Setup: Anker SOLIX X1-H5K-S hybrid (5 kW AC), 15 kWh LFP pack, inverter_hybrid: on.

inverter_limit: 5000
battery_rate_max: 5000
# inverter_loss 0.04 and battery_loss 0.03 (the defaults)

What I see

In the plan the same battery is shown charging and exporting at different rates:

• Charge slot tooltip: "Charging up to 21% at 4.80kW at the import rate ..."
• Export slot tooltip: "Exporting down to 18% at 5.00kW at the export rate ..."

sensor.predbat_inverter_config reports battery_rate_max_charge, battery_rate_max_charge_dc, battery_rate_max_discharge and battery_rate_max_export all at 5 kW, so the configured rates are symmetric.

Where I think the 4.8 comes from

From the v9.6.4 source: get_charge_rate_kw() in output.py caps the charge at inverter_limit * inverter_loss (the GH#5394 change), which is 5.0 x 0.96 = 4.80 kW. The prediction does the same: the SoC rises by 0.388 kWh per 5 minutes in every charge window, which is 4.8 kW x 5/60 x battery_loss 0.97. On the AC side, get_total_inverted() divides charge power by inverter_loss, so 5000 W AC into the inverter plans only 4800 W into the battery.

The export side behaves differently. get_total_inverted() counts battery discharge on the DC side with no loss factor, and the export tooltip uses battery_rate_max_export directly. So export is planned at the full 5.0 kW, while charge is planned at 4.8 kW.

What the hardware does (measured from HA)

Charge, 04:42 to 04:55, flat all the way:

• battery power: -5000 W, within +/-20 W
• grid import: about 5200 W, with a house baseline of 150 to 200 W

Export, 18:42 to 18:55:

• battery power: about 4850 W, wobbling between 4690 and 4900
• grid export: 4570 to 4690 W

So the X1 charges the battery at its full 5000 W, and exports slightly below it. Predbat models it the other way round: it undercharges and overexports. The default loss settings look right for my system. Over a long grid charge I measured about 93 to 95% from grid to stored energy, which is close to inverter_loss 0.04 plus battery_loss 0.03, so I don't want to change those.

What I did and what I'd like to check

To make the charge slot plan at 5.00 kW I raised inverter_limit to 5210 (5210 x 0.96 = 5000). The docs describe inverter_limit as "the maximum AC output power" and say it is "ONLY used by Predbat to improve the quality of the plan".

Questions:

1. Is raising inverter_limit above the nameplate AC rating the intended way to model a battery whose DC charge rate equals the inverter's AC rating? Or is the 4.8 kW intended, because the AC limit is a real ceiling for grid charging and I should leave it?
2. Are there side effects of a larger inverter_limit that I'm missing? I expect it only changes the AC clip applied to PV plus battery, and the charge cap above, and the inverter has delivered up to 5.6 kW briefly for me. Does it also affect discharge or export planning?
3. Is the asymmetry intended, where charge is limited to inverter_limit * inverter_loss but discharge is compared with inverter_limit on the DC side? If inverter_limit is the AC limit, 5 kW out of the battery would be about 4.8 kW at the AC side. If the asymmetry is intended, a one-line note in the docs for inverter_limit would help people who see 4.80 against 5.00 in the plan.
4. Would you rather I used something else, such as inverter_limit_charge, or lowered inverter_loss? I'd avoid lowering inverter_loss, since it changes the economics, and inverter_limit_charge seems to cap rather than raise.

Thanks. Happy to provide a predbat_debug.yaml if useful.

[gcoan]

I can't answer on the asymmetry and why it is what you have found, but this is quite common that inverters will charge and discharge at dissimilar rates from the nominal rate it is placarded as.

You use input_number.predbat_battery_rate_max_scaling and rate_max_scaling_discharge to adjust the charge and discharge rates to match what you get in reality. Mine for example are set to 0.92 and 1.04

[pietrouk]

Thanks for the prompt reply @gcoan

The difference here seems to be on Predbat's planning, not the inverter, see screenshots: