Parallel Runs
Issue #77
resolved
Hi,
Is it possible to run two systems in parallel locked in the same time step rather than in series. I would like to copy some flows from the one model into the other while the analyses progresses
Many thanks
Anton
Comments (4)
-
reporter
-
Anton - I have not attempted the setup of running two models in parallel that talk to each other, but I'm thinking that setting up a custom run in one of them and initializing the other in the same code could allow you to synchronize the runs at a time step level (@ OnIterationConverge() and transfer variables).
Regarding the problem that you are trying to solve, initial implementation of a functionality called "back routing" attempted to address this problem to trigger release decisions based on future delivery conditions, but unfortunately development on that front was suspended some time ago. Some basic back-routing functionality could be enabled in the 'Extensions' in the current version.
-
- changed status to resolved
Question answered.
-
reporter
Hi Enrique
Apologies for the late response - for some reason I missed your reply. Thanks very much for looking at my queries.
What I tried to do at the time was use a spreadsheet to estimate the required releases. With your subsequent help I might have been able to write a function in MODSIM to do the backrouting. My perceived problem with the backrouting was that it took into account the future inflows instead of a conservative estimate of the inflows (which might not be aware of any future rainfall) so it might underestimate the system losses. Following my detailed spreadsheet work the problem went away if one simply put enough balancing storage downstream of the major dams to manage unregulated accruals.
Thanks again for your comments
Anton
- Log in to comment
Maybe I should add more detail on why I was considering two systems. I am trying to model the losses associated with releases made from upstream dams for demands located about 5 days travel time downstream. At the time of the request the downstream irrigators are not aware of future flows and might underestimate the accruals to avoid experiencing shortfalls in supply. This situation is represented by one of the systems which has supressed accruals (say factored by 50%). The other system has actual accruals. The system with the suppressed accruals is used to estimate the release required from the upstream dams, taking into account the suppressed acruals, the water available from some relatively small weirs downstream and lags. This release would then also be applied to the system with the actual accruals to model one timestep as well.
Then the storage in the downstream weir is transferred back from the system with the actual accruals to the system with the supressed accruals to calculate the next days release. The spillage reduction for different downstream weir sizes can then be estimated. .. Any better ideas are welcome ! Suspect I should use a spreadsheet for now ...