And if it is sent every 1min, would it be closer?
So you’re saying modbus with the default of 10s updates, integrated, might be more accurate?
Energy counters are calculated and sent every 15 minutes. Always. You can also see this on VRM, you will see those consumption/solar bars on your overview/dashboard update every 15 minutes. Try it… 
Yes.
Just to be clear, let me use an example to illustrate the challenge. Let’s say you have two energy counters. One is from an MPPT, which says how much PV was produced. The other measures how much energy went from the inverter to loads. The difference between these two then went into the battery, right?
For my example, let’s say I do my calculations once an hour.
Now imagine you have 1kW of sunlight for 30 minutes, and no loads. Then for the next 30 minutes, the PV goes away but you have a 1kW load. Your PV energy counter now shows 0.5kWh produced, and your second energy counter shows 0.5kWh was consumed by loads. Without extra more granular info, you would then assume that the battery did nothing, and the 0.5kWh of PV went straight to the loads. In reality, 0.5kWh was used to charge the battery, and 0.5kWh went from the battery to PV.
Errors in measurements due to assumptions like this reduce if you sample it more often.
You will be surprised how many batteries don’t have energy counters. The BMV has decent counters, but many CAN-bus batteries don’t have them.
Now repeat the above picture with AC-coupled PV on the input of the Multi, an energy meter measuring the grid, and loads on the input side of the Multi. The loads on the input side of the Multi is the difference between the energy measured by the grid meter and the energy measured on the input of the Multi… plus anything the PV-inverter made that is not accounted for elsewhere (so loads must have used it, right?). There is no dedicated measure for consumption, it is calculated…
The same happens with consumption on the output of the Multi where there is also a PV-inverter on the output. The consumption must be what the PV-inverter made, plus anything positive that the Multi supplied on its output, right? In this case, errors are thankfully not to bad, since PV-inverters generally only make energy, they don’t consume it.
1 Like
VRM and HA are pretty accurate on my side all using integration;
Error over 20days;
Consumption 1.5kWh
Grid 0.4kWh
Production 2.4kWh
The battery data does not agree at all though
HA energy panel adds any solar that went to AC loads directly so it never matches, always off by ~15%
Thanks for posting your config @ebendl managed to get my Energy Management working as a result, took me a while to realize that energy distribution animation is cumulative data and only updates hourly, had me scratching my head as to why the consumption was not increasing in real-time 
Was more wondering along the lines of the Advance option, the graphs that can be drilled down into, wanting confirmation that if one sets the upload to 1 min, the core data will be more populated/accurate = more accurate graphs = more accurate 15 min consumption/solar bars.
The PV produced is DC, then conversion losses later it is AC. So you could implement a type of conversion loss estimate.
It varies, according to power level, and if the fans are on or not. But 88% is not a bad estimate to use if you’re going with a static number. You could also make your own linear interpolation function by taking a few of your own data points
@ebendl Do you find that this works for you?
Reason being, I pretty much copy pastaed all of your sensors and pulled it into HA Energy which works great, I just have a consumed battery out spike that happens from time to time which doesn’t match what’s actually happening on the battery.