openego / eDisGo / 36104705703

84%

push

github

Weight line impedance by num_parallel when reducing the grid (#768)

* fix(tools): weight line impedance by num_parallel when reducing

A line that stands for several parallel systems came out of
spatial_complexity_reduction with num_parallel times its impedance.

aggregate_lines_df rebuilds r and x from the per-km values of the line type,
combining the grouped lines in parallel. Those values describe ONE system, and
num_parallel was never part of the calculation -- while num_parallel and s_nom
were carried over intact, so nothing in the reduced grid showed that its
impedance no longer matched its own metadata.

Few lines carry parallel systems, but they sit where a feeder leaves the
station and are therefore on the path of everything behind them. On MV grid
32064 an eight-system feeder came out at 0.164 Ohm/km instead of 0.0205 -- the
catalogue value of NA2XS2Y 3x1x185, i.e. exactly eight times too resistive.
Measured over that grid: all 34 one-to-one mapped lines with num_parallel > 1
had their resistance per km multiplied by exactly num_parallel, the
load-weighted path resistance from the station to the loads doubled from 0.458
to 0.940, 81.5 % of loads ended up electrically further away, and the voltage
band went from 0.908-1.033 to 0.818-1.136. reinforce() then aborted with

    MaximumIterationError: Over-voltage issues for the following nodes in MV
    topology could not be solved

on a grid that needs no reinforcement at all before the reduction. (The
message says over-voltage; the buses it names sit at 0.926 pu.)

Each line now counts as num_parallel systems of its type, so the per-km
admittance is num_parallel / value_per_km. With that, the 35 affected lines of
2421 keep their resistance per km exactly, the voltage band comes back to
0.895-1.040 and the expansion runs through with 25 measures.

The test reduces the same grid twice, once with every line as a single system
and once with n, and asserts the impedances differ by exactly n. num_pa... (continued)

5 of 5 new or added lines in 1 file covered. (100.0%)

9558 of 11379 relevant lines covered (84.0%)

0.84 hits per line