[NEMO-devel] Polygons in NEMO and optimization displayed results
bje at air.net.au
Thu Jun 29 17:35:04 AEST 2017
> My question is about the polygons. Is it like a compulsory to define
> polygons, because I saw in your first two publications, you didn't
> mention about the polygons at all, or in other words, can we run
> NEMO without having the predefined polygons?
Earlier work used a concept of regions only, for siting generators and
demands. This was extended to a finer spatial resolution of polygons
(as in the AEMO 100% RE study). There must be at least one polygon
and each polygon must belong to a region. For your initial work, you
can use the polygon number `polygons.wildcard' which makes everything
happy. In the long run, though, you will probably want to modify
polygons.py and regions.py to localise them to your area.
> About the displayed results, is it the last generations (in this
> case after 5th generations) to be displayed in NEMO? Can we also
> displayed other generations result?
Good question. Your output provides a good example. The fitness
column shows the fittest individual in that generation. The
"hallfame" shows the fittest individual ever seen. Note that in
generation 4 below, that the fittest individual (129.992) is higher
than the all-time best. We do not take the fittest individual in the
final generation, but the fittest individual overall.
nevals min min
0 19 123.6 123.6
1 19 127.99 123.6
2 19 122.026 122.026
3 19 115.008 115.008
4 19 129.992 115.008
You can use the --trace-file option (see documentation) to write out a
CSV file containing every candidate tried (in the example above, that
would be 5 x 19) and their fitness.
More information about the nemo-devel