Right, sense is not accurate. There are only two reasons I can think of that would result in such a problem. The first is calibration of the voltage and current measurements. I have no idea how the calibration is done, so I can’t comment. The only other reason is if it stops measuring the voltage and current for periods of time, so the total is recorded low and incorrectly. Certainly, I also see negative 8-10% errors, which are very high, so I doubt it is a calibration issue. On my system, I have three measurements being made of my solar production. The first is the electricity company’s meter, the second is the meter on my inverters. These two agree exactly within the precision of the numbers on my electricity bill. The third measurement is the Sense numbers, which are not reliable, with most days being different from the report from my inverters. I have 40 micro-inverters, which report to their own base station to produce the total solar production.
I agree that my Sense reports numbers that are always lower than the actual production, so it is a one way error that can be explained by dropouts, and I have not been able to come up with another explanation. Clearly, this is an accuracy problem, not a precision problem.
Also, I agree that there is a small offset at night. I think “recording the inverter power” means that Sense is likely recording the quiescent power used by the inverter(s). I do not know whether or not the number is being subtracted from the total production numbers provided by Sense’s reports, but it is not being baseline subtracted from Sense’s meter.
As for living with it; its wrong, plain and simple. It means that all the Sense numbers may well be out one way or another, so how can you rely on reported appliance power and energy usage to make decisions? While sense is a great device, and addictive, it has flaws too, including misidentification of appliances when they turn on and off. Again, this is a “10%” problem, but it is there nevertheless.