Spilled Energy

[ThanosGkou]

Hi and sorry for the long message.

I am doing some expansion planning with Antares (Simulator 8.2.2., Xpansion 0.8).

I am solving with "expansion_accurate" setting, and let's say zero day-ahead reserves for simplification. I have set a high cost of spilled energy for the "parent" area, and a zero cost of spilled energy for the virtual candidate areas (so as not to penalize un-built capacity).

The issue is that the Xpansion seems to over-build RES units. In the economy output of the expansion output folder, the spilled energy is reported as zero. But when I re-run a "state" simulation (after integrating the expansion results to a new "state" model), there is quite often spilled energy. ("State" simulation, I mean a normal AntaresSimulator run, with the updated study)

I thought that perhaps, the benefit from the decreased operational cost of having more RES, overwhelms the occasional high cost of spilled energy. But this is not the case. If I manually decrease "a bit" the expansion-dictated RES, there is a sweet spot where the overall system cost is lower as compared to the exact expansion results.

Has this been observed again? Is there something I could be missing?

Remark: Instead of using the automatically updated study (arbitrarily large capacity in the candidate areas, bottlenecked by line capacities towards the parent area), I use an "Invested" area, in which I manually (scripting actually) integrate the results of the expansion. This "Invested" area is originally empty, so no interaction with expansion module. Then, I integrate the expansion results in it. The "Invested" area is set with a high cost of spilled energy. Whereas, if one uses the automatically updated study and does not set a high cost of spilled energy to the candidate areas, this may not be noticed as a spilled energy in the parent area, but rather, in the candidate area.

[tbittar]

Thank you for your question !

It is hard to give a precise answer without looking at the study.

As your question is related to a study case and not to a precise technical bug of Xpansion, you may contact antares-simulator-services@rte-international.com to get support.

The only idea that comes to my mind is that Xpansion uses a relaxed version of the Antares problem during the optimization. This can lead to differences in the results when reusing the solution of Xpansion back in the Antares study.

[ThanosGkou]

Hi, thanks for your answer.

Just summarizing some remarks:

• (Specifically) When there is a large cost differential between current-fleet and candidate-fleet (e.g. high-costs gas-fired units vs low annuity-RES candidates), the expansion solver often builds too much RES*.

(below, when I refer to candidates, I am referring to non-dispatchable, RES candidates) *What is meant by too much RES

• If one makes a variant study, where all candidate areas are deactivated, and takes the optimal capacities from the expansion output, and plugs them directly into RES clusters, there is very often spilled energy (sometimes significant) alternatively:
• If one takes the updated expansion study (where link capacities = link-profile * optimal capacity), and runs it as an Economy simulation: The actual flow b etween candidate areas and the "parent" area is not always equal to the line's capacity: There is some dispatching in the candidate area, even though it is non-dispatchable.

Just in case there's time to review it, I attach two studies, indicating this effect. (without the solutions, for uploading issues). There are fully commented in "user notes", also attaching a txt with a description of the setup.

description.txt without_solutions.zip