Documentation inconsistent in modeling multi-phase site meters

[davidwe]

What should be assigned to submeter_of if you have a 3-phase main meter and you know to which specific line the submeters are connected?

The 3-phase main meter is modeled as the first three entries in the dict elec_meters, which are all site_meters.

• In the tutorial (Chapter 1.1.1 Simple example in nilm_metadata/docs/source/tutorial.rst) every submeter is submeter_of one of this main meters, so either 1 or 2.
• But in the documentation of ElecMeter (Chapter 2.4 in nilm_metadata/docs/source/dataset_metadata.rst) it says (even twice in a paragraph) "All non-site-meters directly downstream of the site meters should set submeter_of=0. Optionally also use phase to describe which phase this meter measures." In the metadata for REDD it's done this way, too, but without specifying the phase field.

So which is the right way to do it?

[JackKelly]

If you know the specific phase each submeter is connected to then please specify the precise site meter. (i.e. option 1 in your list)

Please only use submeter_of=0 if you don't know which exact site meter is upstream. This is the case in REDD: we don't know which of the two site meters is upstream of each submeter.

[davidwe]

Okay, that is also the way I have done it. Thanks for the quick response!

I was asking that because I get RuntimeError: Tree has more than one root! when calling for example elec.describe() from nilmtk on my dataset. Do I have to read that error as "not implemented yet" or "you modeled your electric installation wrong"? Do I maybe have to split the network into three subtrees manually?