Solar-powered inverters in OpenDTU-OnBattery and PID to compensate inaccuracies

[schlimmchen]

Originally by @mathieucarbou in tbnobody#2511 (comment):

For the PID, its use does not relate to the required accuracy, or to speedup the limit. It more relates to ability to react more properly to an inaccurate correction. For example, if I set a target excess to -300W, and I need to set the limit to 300W on 10 inverters to arrive to this target, in reality, one inverter will produce 280W, one 330W, one will be under some shade, etc. So the PID will be able to auto-correct such inaccurate adjustment by recalculating a new correction to apply, until the next measurement comes. A PID would also be more tolerant and able to correct the uneven production of inverters, caused by shade, difference of power, orientation, etc.

without PID, if you simply correct by computing a difference and spreading the new limit on inverters, the reality is that the real correction won't be as expected by the calculation because of too many unpredictable factors.

But I am pretty sure that in your code, to do that, you are using some kind of PID, but just more specific, with the Kp term only (which is a correction done based on a percentage of the error == target - current measurement).

The graph that you have is very good I think!
I know that it takes a few seconds for inverters to adhere to a new limit. So doing that in 5 seconds is nice!

The Kp Ki Kd terms of a PID allows also to tweak its responsiveness.
In your graph example, you seem to have quite stable home consumption after the boiler is disabled.
But imagine that you had someone cooking in the kitchen ;-) You would get a lot of on/off with different appliances. So a PID is also able to "slowdown" to reach a target by smoothing its corrections to avoid these sort of abrupt correction (coming a few seconds later in effect) in the case of a sequence of lot of high / low consumptions. It does that thanks to his integral term and summations.

[schlimmchen]

I enjoy you input and the discussion, but since I would not want so much off-topic discussion in issues in my projects, I am taking this here 😉

Yeah, now I get it! The PID you suggest would help "carry" the limit to a value that drags the effective AC output to the actually expected value.

@AndreasBoehm integrated a rather sophisticated compensation scheme for shaded modules and varying input power.

However, in the future, we expect to offload all of this to the newer firmware versions for the HMT and HMS inverters (also HM? not sure), which does all this automatically/internally. The limit set on the inverter is then equal to the target AC output power. If the inputs provide varying amount of power, the inverter will compensate internally.

But I am pretty sure that in your code, to do that, you are using some kind of PID, but just more specific, with the Kp term only (which is a correction done based on a percentage of the error == target - current measurement).

I also think that I understand what you mean by this. And that would be true.

The graph that you have is very good I think!

Happy to hear that! I would still like to improve on this, but it is getting tricky.

I would love to have you use OpenDTU-OnBattery. Don't hesitate to ask questions.

[mathieucarbou]

Thanks! I will keep this tab opened :-)