Accurate or not?

My system consists of an Arduino One with an EmonTx Shield (V2.5) and an EmonEsp. Here is a picture of my installation.

It’s 1 phase. Each of the branches is called a ‘group’. I have 2 clamps: CT1 measures the total current and CT2 the current of one group. This particular group has only 1 wall outlet, and I installed a Shelly wall plug. At the moment I measure everything in Home Assistant.

The results from Emon and Shelly look more or less the same (except at low power).

However: the other clamp measuring the total power seems way off. I have measured the energy in kWh. The energy both comes directrly from EmonTx (using the Continuous Monitoring lib) and from Home Assistant (using the Riemann sum integration) and these 2 are nearly similar.

But over some time I measured 34 kWh while my old Feraris meter showed 53 kWh. This is a very big difference. What could be wrong?

EmonTx reports about 200V Vrms, while I measure 235 V with a multimeter. On the 9V AC adapter I measure 9.5 V. I don’t think this can explain the big difference, can it?

If my ferarismeter is wrong, I have been paying way to much for my electricity over the past years.

No, but it goes a long way towards it.

Does your multimeter measure a.c. current with reasonable accuracy, and have you calibrated your emonTx Shield using it? The default calibration in the sketch assumes the UK a.c. adapter, I can’t see yours in the picture, but if you have the 77DE-06-09-MI, the calibration is wrong for that. But I think there’s another problem with your adapter - the output is specified as 11.5 V ±5% with a 230 V input on no load (which is effectively the case), so yours is way out of specification at 9.5 V out with 235 V in. Correcting this alone makes your 34 kW come to 42 kW. That explains half the difference. You need to adjust the voltage calibration constant in your sketch so that the software reports the same voltage as your multimeter.
Then you need to calibrate the current measurement. Ideally, do this at the sort of current where you consume the most energy, if that’s above the current your multimeter can handle, just use two or more turns through the c.t. to multiply the current that the c.t. sees. Adjust the current calibration constant so that the sketch reports the same current (or the correct multiple) as your meter. Finally, use a pure resistive load (e.g. an electric kettle, or an electric heater without a fan) and adjust the phase calibration to give as close to unity power factor (real power = apparent power) as possible. Do this for both c.t’s.

Then check against your Ferraris meter. You should be within about 1% plus the accuracy of your multimeter on a.c. voltage plus the accuracy of your multimeter on a.c. current. (So overall, within about 5%).

That is unlikely, but not unknown. I don’t know the conditions attached to your supply, but in the UK, I can ask for my meter to be tested. If it’s out of specification, the company pays for the test and adjusts the bill (either up or down!). If it is within specification, I have to pay for the work in taking the meter out etc, even though the test is free.