Emonpi over reading power1 after power cycle

Pulling my hair out with this, hopefully someone can help.

At the weekend, the emonpi was unplugged (without the right shutdown sequence). Since plugging back in, the power1 input is consistently reading 500w too high.

I have battery and solar, and the battery usually does a very good job of ensuring there is only a trickle coming from the grid.

Two screenshots attached. The first from before the power cycle; and the second is from today. You can see the red area is consistently 500w above the rest indicating I’m pulling 500w from the grid despite me the solar/battery matching the load correctly.

Any ideas what could be causing this?

An additional data play where you can see there is suddenly a distinct gap between use and import at a time when the battery is schedule to not discharge to load.

Did the emonPi update itself when you started it up after the interruption?

If this is all that’s happened, it suggests either you had some calibration data not permanently saved, or a calibration value or a processing step has been corrupted. But this would be most unusual because there’s no general mechanism, other than in the front end sketch or as part of the input processing in emonCMS, to add a constant value to an input. Think in terms of y = mx + c, it’s only possible to add the ‘c’ term in the sketch or in emonCMS.

Are you sure it is a constant difference? Are you still measuring what you intend and think you are?
Have you checked your processing steps in emonCMS, for both inputs? Was there something else going into the maths that’s gone away?

I’m afraid this is all I can suggest, given what you’re written.

Thanks Robert

I’ve learnt something today. Turns out the wife unplugged the AC voltage reference adaptor, not the 5v power supply.

I shutdown the Pi using the normal procedure, unplugged both power supplies, plugged in the voltage reference transformer first, then the 5v supply and let it boot up.

Problem is now resolved. It had obviously lost the VRMS defence to give an accurate solar reading that was impacting any other input that depended on that feed.