Experimental Load tracker

[Javert1969]

I did set up the experimental load tracker that you recommended.

I have noted today that the experimental tracker is clocking higher than the existing integration stats.
Experimental 19.91
Alpha integration load so far today 19.26
Alpha app / website 18.77

(this includes about 9KW of car charging FYI)

I am not sure whether this is a timing difference, or a rounding difference, or if not, which one is more accurate to the real situation. Did you find anything similar?

[PlainSeer]

Yes, I have seen differences between the derived sensor, the AlphaESS integration and the AlphaESS cloud figures. I wouldn’t assume from one partial day that any one of them is definitely wrong.

The experimental sensor integrates the live house-load power readings reported through Modbus. Its result therefore depends on the update frequency, the integration method and exactly what the source power sensor includes. The AlphaESS integration and cloud figures may use separate inverter counters, different measurement boundaries, rounding or delayed cloud updates.

Your current figures are reasonably close:

• Experimental: 19.91 kWh
• Integration: 19.26 kWh
• AlphaESS app: 18.77 kWh

The experimental value is about 3.4% above the integration and around 6% above the app. The 9 kWh of relatively steady EV charging may make any small scaling or timing difference easier to see.

I would compare them over several complete midnight-to-midnight days rather than trying to calibrate from today’s running totals. Record all three at approximately the same time after allowing the cloud figure to update, and check that each resets at the expected time. It would also be useful to compare a day with substantial EV charging against a day without it.

As a broad cross-check, the house-load total should make sense against the day’s PV, grid import/export and battery charge/discharge energy, allowing for conversion losses. That will give a better indication of which figure is closest to the actual energy balance than comparing the three load totals alone.

If the experimental sensor remains consistently high by a similar percentage over several complete days, that may indicate a systematic difference in what the live Modbus sensor measures. If the gap varies significantly, update timing and sampling are more likely explanations.

For the moment I would leave it unchanged, collect a few complete days and avoid applying an offset until there is a repeatable pattern.