search

4flixt / 2019_WNTR_Surrogate_Model

5 stars 2 forks source link

Stepwise simulation (with resimulation of the past): Quality is not consistent #22

Closed 4flixt closed 5 years ago

4flixt commented 5 years ago

I investigated the stepwise simulation to make sure that results don't differ. Unfortunately, this seems not to be the case for the quality at the junctions. I tested:

with open('./results_sim_time448.pkl', 'rb') as f:
    results_1 = pickle.load(f)

with open('./results_sim_time447.pkl', 'rb') as f:
    results_2 = pickle.load(f)
(results_1.node['quality']-results_2.node['quality']).plot(legend=False)

Which yielded: image

The difference is only 3600s which is not much in total but I wonder where this comes from.

I also tested for the nodes:

and for the links links:

These quantities are fine. This also means, that I dont have a problem, yet. Since I am only doing hydraulics so far.

acominola commented 5 years ago

Felix, where do I find these results?

4flixt commented 5 years ago

Sorry I haven't uploaded them. You'll find some attached (I deleted most of the intermediate timesteps to keep space free. Here are some that were left by chance and I hope the bug shows there as well).

results_sim_time71.zip

acominola commented 5 years ago

I checked this for the results you sent me and then ran an experiment with a simulation horizon of 500 hours and I always get that the sum of (results_1.node['quality']-results_2.node['quality']) is 0.0. Not sure whether something was wrong with overwriting in your case

acominola commented 5 years ago

(Note that I have done this experiment with my original run_simulation_1stepControl.py script, not with following versions)

4flixt commented 5 years ago

Ok and in any case the difference was only macimum one timestep. Just wanted to make sure there was no underlying problem.

Andrea notifications@github.com schrieb am Do. 14. Nov. 2019 um 23:04:

(Note that I have done this experiment with my original run_simulation_1stepControl.py script, not with following versions)

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/4flixt/2019_WNTR_Surrogate_Model/issues/22?email_source=notifications&email_token=AH2XOC6LT2RA3XNHH5HXXO3QTXDOLA5CNFSM4JNO75LKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEEDOXUQ#issuecomment-554101714, or unsubscribe https://github.com/notifications/unsubscribe-auth/AH2XOC53M6TNYL27JKVXS7TQTXDOLANCNFSM4JNO75LA .