Support negative start time in EnergyPlus

[mwetter]

The following command

simulateModel("Buildings.ThermalZones.EnergyPlus.Validation.OneZone", startTime=-864000, stopTime=864000, method="Cvode", tolerance=1e-06, resultFile="OneZone");

produces

[kbenne]

What are the outside conditions when a simulation is started in a scenario like this with a negative start time? I need to correlate with a time in the weather file.

[mwetter]

If the E+ weather file is from t=0 to t=365 days, then you can look up the outside conditions at t-365 days.

[kbenne]

Yikes!

[mwetter]

In FY21 AOP for NREL.

[mwetter]

Todo for the release:

• [x] On the Modelica side, issue an error if the simulation time is negative.

[mwetter]

@kbenne : This issue is not yet fixed in the latest build that is merged through https://github.com/lbl-srg/modelica-buildings/pull/2984

Commit 06cc58dc25 on branch issue1938_spawn_negative_startTime adds the validation case Buildings/Resources/Scripts/Dymola/ThermalZones/EnergyPlus_9_6_0/Validation/RunPeriod/NegativeStartTime.mos which produces the following output:

[kbenne]

Hmm. That's interesting. I can't perceive why it is doing this, although the testing I'm doing at my level doesn't attempt actual simulation. Take a look at the unit test in this commit.

I'll have to dig a bit deeper.

[mwetter]

You may need to check if E+ does internally the right thing, for example by verifying that it picks the same weather data for t=-1 days as it does for t=364 days; or whether the simulations for these two day specifications produce the same indoor temperatures.

[kbenne]

Yeah I certainly have some investigation to do. But the whole point of the strategy was that EnergyPlus would never actually see a negative time. I probably need to rig up a test that actually does simulation for day 364 and day -1, which will be a bit more involved.

Note how this so closely relates to https://github.com/lbl-srg/modelica-buildings/issues/2453 because simulating day 364 actually requires EnergyPlus to simulate through the first 363 days to arrive at the start time.

[mwetter]

Delay for later, not part of release 9.0.

Problem is that if simulation starts at -1 day, then week-day of Jan 1 must be the same as if it were to start on Jan 1.