[BUG] Zero-sequence modelling of three-winding transformer

[TonyXiang8787]

We have two issues which need fix for three-winding transformer

Separate attributes for zero-sequence magnetization

In #1197, the zero-sequence modelling of two-winding transformer has been improved. User can now specify zero-sequence magnetization current/loss to give detailed zero-sequence modelling. This is espacially neede for Yyn tranformers.

We need to implement similar things for three-winding transformer.

Disappeared delta-winding

In the modelling we divide three-winding transformer to three two-winding tranformers with an aritificial node. We always use node_1 as the aritificial YNyn transformer. This is correct for positive and negative sequence. But for zero-sequence the current implementation is not correct.

Given a Y/yn/delta transformer, the equivalant modelling will be.

node_1 ---YN/yn---- artificial_node ---Y/yn-----  node_2
                             |
                             |
                          Y/delta--- node_3

In this modelling, node_2 will not have a delta winding at primary side to allow zero-sequence current to circulate. However, this is not the reality. All three windings are bundle together in the iron core, and the delta-winding from node_3 should be able to provide proper zero-sequence circulation.

The model has to be adjusted to not to blindly use node_1 to create the aritificial YNyn transformer. If there is a delta-winding in the configuration, the aritificial YNyn transformer should be created from the delta-winding side.