Transformer with voltage regulator about gridcal HOT 14 CLOSED

Hi Michael,

Excellent work. I have to think about how to integrate this functionality. As it is formulated (using the tap changer objects) I'd implement it as a function inside the class PowerFlowMP, along with the PV-PQ switching. Both the the tap changing and PQ-PV switching logic are discrete operations that belong to the so-called outer loop.

I'll do this as soon as I get some time (this week probably). You're welcome to do it yourself if you need this sooner.

from gridcal.

I'd like to do it, I'm in no particular hurry (especially since this works in the meantime). I'll have a deeper look at the outer loop and try to integrate it gracefully.

About the other points, could you please confirm that:

1. GridCal development must be compatible with Python 3.5 (i.e. f-strings not allowed);
2. I may include tests meant to be run with pytest in the tutorials.

Thanks.

from gridcal.

I modified the branch to include the two variables that you needed for tap control since that part was trickier (I found a bug on the process too) and I copied the tap control to a function in PowerFlowMP (it does nothing, just copied there)

I confirm that I require python 3.5 or higher no problem with that, and I'll include pytest in the requirements of the setup file.

from gridcal.

Ok thanks.

About Python 3.5, I meant: Must all code be backward compatible with version 3.5? If I include f-strings in my code, it won't run under 3.5 and before. If you're developping on 3.6 or 3.7, I would change the minimum version to 3.6 in the README.md to allow f-strings.

from gridcal.

Ok, no problem, I'll set it to python 3.6

from gridcal.

Great.

I'm working on the pull request, I'm having minor trouble getting it to work with a fork and filtering unwanted stuff. But it works and shouldn't add any significant calculation/compilation if now voltage regulators are used.

Edit: Ok, I think I've got it. Here's the PR: #24

from gridcal.

Simply reopened it until it's fully resolved. Will close my PR #24.

from gridcal.

Hi Santiago,

Did you October 5 commits integrate this feature? Is it ready for testing?

Thank you.

from gridcal.

Hi Michael,

It's been quite hectic these days.

I have just implemented the feature in a thread safe way, but it is not tested at all.
If you could see if my implementation follows your initial design it'll be nice.

For your reference this is how the compilation works:

MultiCircuit.compile() -> Gathers the properties from the objects and creates a NumericalCircuit instance.

NumericalCircuit.compute() -> Computes the admittance matrix, Sbus, Vbus, computes the islands and returns a list of calculation-ready islands (instances of CalculationInputs)

The instances of CalculationInputs are used to run any simulation ensuring that the circuit that the internal matrices and vectores represent is not singular.

So, if the power flow needs to control the transformer taps, I need to copy the information from the objects all the way to the CalculationInputs. This produces a thread-safe object that can be used in parallel computation. It has taken me 3 years to figure this architecture out and I have changed GridCal's internal engine 2 times because of this. So this last one seems to be the definitive architecture since it is 13 times faster than the last and it is all vector-oriented, and it allows three-phase calculations with minimal changes (I think)

So, if you need any clarification or you find any bug please let me know.

from gridcal.

I sent 3x PR to fix a few bugs that prevented it from working in my test. With these all my tests pass! Thanks for taking the time to integrate this feature, I can see it wasn't exactly easy.

from gridcal.

I added a comment on commit 848f813

The way control_max_iter is calculated won't produce the desired effect. I suggested alternatives.

Thanks.

from gridcal.

BTW,

I am implementing a storage active power dispatch to control the adjacent branches overloading.

I am not sure how stupid this idea is. Have you ever seen anything similar?

BR,
Santiago

from gridcal.

Hi,

It's certainly not stupid, I currently have a project for a customer who could beneficiate from this, and I expect others to follow as regulation adapts to energy storage. For example, say we have a relatively constant 20 MW generation and a variable 20 MW consumption tied to the same POI (point of interconnection), and very different buying and selling energy costs. Say the buying cost is ~10 times greater than the selling cost.

Ideally, we want the generation and consumption to be always equal, so that no energy transits through the POI. In order to acheive this, an active storage solution could be used as a buffer to "clip" generation to the consumption, and to compensate for the lack of generation when consumption is higher than generation.

That's only one of probably many applications, but it would definitely be a nice feature to incorporate. I could also be used in larger utility studies, i.e. not at the individual facility level. I would probably start by having a look at how PyPSA did it, I haven't played with it much but they seem to have already integrated this feature.

They have examples in which they use energy storage to reduce branches overloading at the utility level, as you mentioned. That would be another worthy application.

Are you going to create a new development branch for this?

from gridcal.

Hi,

I already have the storage dispatch feature implemented in the optimal power flow in DC and AC approximations. This is how PyPSA does it (I am quite sure)

The thing is that I am with a real distribution grid from a EU project where the implemented OPF algorithms are not doing as I expect, whereas in IEEE30 it does exactly what I'd expect. So I want a "rule-of-thumb" solution that works without caring about optimality. In the past I did this by implementing a truncable Nelder-Mead optimization.

I'll include a rough test as a power flow option in the master branch. Since it can be disables (and is by default) there's no need of a new branch.

from gridcal.