Accounting for charging loss seen in load

[DFEvans]

My inverter reports additional load during grid charging, which I presume is a representation of the conversion losses when charging from the grid. Batpred observes these, and appears to treats them as part of the normal load - the two increased bits of overnight load I've indicated below:

I've configured it to average multiple days and once it has more data it should improve. However, I can also imagine that it will be routinely charging from the grid at similar times most days, and will possibly then double count the conversion losses when modelling - including it what it believes is the historic load, plus an extra contribution from its predicted conversion losses.

Does Batpred have any functionality to understand these losses? It's a minor correction, but adds up to a possible routine ~3kWh/day of overpredicted load for my system if doing full charges from grid.

[gcoan]

The predbat losses should be able to give you a good representation of how your inverter operates https://springfall2008.github.io/batpred/customisation/#battery-loss-options

If you have extra losses from grid charging then this sounds like you need to increase your inverter loss which is applied when converting from DC to AC or vice versa, versus solar charging where only the battery loss is applied.

Predbat uses these losses to predict what your battery SoC will be over the future plan, taking into account predicted solar generation and house load.

[DFEvans]

Ah, it's just occurred to me why it didn't handle this - I presume predbat only corrects for charging that it commands, and won't correct for charging commanded by another system when in Read Only mode. That seems fairly obvious in retrospect!

[gcoan]

I wasn't sure I understood your reply at first, so went back to re-read your original question:

My inverter reports additional load during grid charging, which I presume is a representation of the conversion losses when charging from the grid

So if I've understood your inverter behaviour properly, and taking a silly example to illustrate it, your house load is static at 1kWh. You then start charging your battery at 2kWh and your inverter's conversion losses are 50%, so the 2kWh grid import results in 1kWh in the battery

And your inverter records that 1kWh lost in conversion as being part of house load, so your house load is artificially high when charging the battery.

Is that a fair summary?

Your conclusion of how predbat works isn't quite right unfortunately, it has no real knowledge of what has happened in the past, and certainly no idea of what commands it issued and the affects of those commands.

Principally Predbat takes the house load history and (via days_previous) will work out the average house load for each 30 minute slot of the forward plan.
It then overlays solar generation to those slots.
Then starting from the current SoC kWh it projects what will happen to SoC kWh, rises with solar generation, falls with house load. Inverter and battery losses are included in that calculation as I linked earlier.

Then predbat starts cost optimisation to work out whether it should charge the battery or force discharge the battery across the forward plan, applying inverter and battery losses to calculate the SoC kWh across time, and then optimising import, export and battery hold activities to come up with the cheapest overall electricity price for you.

That's a brief summary of how it works.

So if your house load is over-inflated because your inverter includes conversion losses in the house load figure, then there's not a lot Predbat can do about this because it relies on the house load figure as a key input to the plan projection.

You could create your own custom house load sensor, and I do that myself as I have two inverters that share the load so neither has a full picture of load. The code is in the documentation but in essence house load = import - export + battery discharge - battery charge + solar generation

This works great but it does mean that any battery conversion losses get lumped in as part of house load which is what your inverter does. I just live with it. I have days_previous set to days 2 through 8, I set modal option on so the lowest day is discarded, and worst case I have a slightly over-estimated house load figure but it doesn't materially break the plan for me

[DFEvans]

Yes, I think you're explaining the situation I have. Taking some numbers from overnight when load is stable at 0.5kW:

• Discharging from battery, Battery Power = -0.5, Load = 0.5, Grid = 0.0
• Charging battery from grid, Battery Power = 11.5, Load = 1.5, Grid = 13.0
• Using grid to feed load (freeze charge), Battery Power = 0.0, Load = 0.5, Grid = 0.5

So when charging from the grid, there is an extra load of ~1kW reported to ensure that the reported power values balance at 0.

I suspect you're right that it's probably not a material difference - the 3 kWh per day I might expect as a worst case is within day to day power use variations. Predbat's use of hold/freeze charge modes that aim at a target SoC (not charging time) and its ability to react to the situation in real time also mean it should happily deal with a difference in change time of 10-20 minutes - usually in it's favour.

[gcoan]

I think it is within the margins of error and as you say predbat recalculates and calibrates the plan every 10 minutes so material differences can be adjusted for.

Thanks for the real world example.

You could make a template sensor for house load that takes the inverter load sensor and adjusts it for the inverter charging loss when the inverter is charging if need be?

[DFEvans]

Thanks for the thoughts - yes, such a sensor might be something to consider.

(Also, apologies for poor threading ettiquette!)