My first guess is you’re not measuring what you think you’re measuring. The second is you’re not doing the additions and subtractions that you think you’re doing.
So what is driving each of those dials? Do you have a.c. adapters so you are measuring real power, or are you assuming a voltage and using apparent power? I think the difference is too great to be a calibration issue, but have you checked that?
OK, you have an immediate problem with the emonTx “power”. It won’t be accurate. Without an a.c. adapter, it assumes 230 V, even thought the UK ‘centre’ voltage is 240 V. Your error there, if the current is correct, can be ±10%. I’d expect it to be reading 4-5% low, which makes your problem worse. You can improve its accuracy by, in emoncms, dividing the power by 230 (multiply by 0.004348) and multiplying by the Pi’s Vrms.
I still don’t know how you get those feeds. Personally, I would have put a Pi c.t. on the grid infeed and used a centre-zero dial to show nett import/export, then the second Pi c.t. on the solar infeed. Generally, because there will be a variation in the sensitivity between all your inputs (unless you’ve calibrated very carefully), it’s a bad idea to derive a feed by subtracting two nearly equal quantities. So I wouldn’t put a c.t. on the house side of where the solar infeed joins, to measure use directly, and then derive the nett grid power by subtraction, because by design your immersion controller aims to make that near-zero when it can (and Robin’s diverter would make it exactly zero).
I’m guessing the controller uses phase angle control, the emonTx will measure the rms current of a load like that reasonably accurately, so I don’t anticipate that being a problem.
PSU added to emontx, inputs tweaked.
now will wait till we get some sun tomorrow to see the results.
Will post results that I hope will be much better.
thanks again for assistance