Repository navigation
Replies: 1 comment 16 replies
|
Not subtracting energy already consumed from the forecasted usage is indeed an issue making the charging start to late, there is however another one: household energy usage is not evenly distributed over the day. I'm our household we follow the instruction to run energy intensive equipment as much as possible during the time when there is a surplus of solar energy, meaning that the surplus energy left to charge the battery, that omnibattery calculated as solar minus the average daily usage is in fact much lower. So far every day I tried to use predictive charging, this resulted in not getting my battery charged sufficiently to not run my battery empty through the night. Even with a 4hr safety margin! So any profit made of exporting excess solar in the morning was completely undone -and more- by energy I had to import during the night after. I have therefor disabled this feature. And asked for a better feature in #237 |
Uh oh!
There was an error while loading. Please reload this page.
Uh oh!
There was an error while loading. Please reload this page.
What do you need help with?
Automations or energy management
What are you trying to do?
I use variable charge and discharge values which can vary throughout the day depending on my energy mix. This leads to the predictive charging system reevaluating charging windows several times at one day.
One thing I noticed is that when per example the predictive charging systems estimates the grid charging at 12pm it always takes the full expected household consumption (7 day average) into account. That is not correct in my opinion because the household already used some that estimated energy (e.g. 1,2 kWh of estimated 5,8 kWh per day already used till 12pm). Therefore Omnibattery predicts way too high grid charging demand at the time it starts to charge because it estimated way too high household consumption.
For me specifically this leads to too early grid charging resulting in a battery full of grid energy while there would be plenty of solar energy left later at that day (my solar system is facing west). As said the grid demand is calculated too high which also leads to the charging delay not triggering (the charging delta is too high and I won't set it to unnaturally high values because it would delay way too hard when demand and solar prediction is even).
Do I have a thinking error here in the logic of the integration (because it tries to do rolling 24h blocks) or is it necessary to cut the already used household energy at that day from the estimated average household need?
What is happening instead?
The whole estimated household energy (7 day average) is always used at full for demand management. The estimation of needed energy is too static for a changing system over the day which leads to over-/undercharging.
Imo the predictive charging is a great feature which needs a bit of work and a more dynamic approach to be truly amazing.
Your setup
Omnibattery version: 1.2
Home Assistant version: 2026.08
Battery brand and exact model: Marstek Venus C
Battery firmware version: I don't know
Connection method or adapter: Modbus
What have you already tried?
Checked the sensor binary_sensor.marstek_venus_system_predictive_charging_active which should already hold the already used household energy in attribute household_consumption_battery_window_kwh?
Checked the documentation for the predictive charging logic.
Relevant logs or screenshots
No response
Confirmations
All reactions