Appendix g dev debug unmet load hour issues

[weilixu]

Bug fixes including

1. Add increment values that breaks the unmet load hour break loop
2. Add statement to close SQL sessions before running a new simulation in the same folder (issues under Window OS)
3. Add error handler when simulation failed in the unmet load hour loop.
4. Add schedule type check for optimal start EMS code

[mdahlhausen]

@weilixu @lymereJ 8 is a lot of adjustments; that's a lot of time a user will be at their PC while the measure is looping through sizing runs. I have a commit drafted to set it to three instead. Is that ok?

[lymereJ]

@weilixu @lymereJ 8 is a lot of adjustments; that's a lot of time a user will be at their PC while the measure is looping through sizing runs. I have a commit drafted to set it to three instead. Is that ok?

@dmaddoxwhite, is that okay?

[weilixu]

s; that's a lot of time a user will be at their PC while the measure is looping through sizing runs. I have a commit drafted to set it to three instead. Is that ok?

I am ok with it, I wonder if we should bump up the multiplier to account for the reduced iterations.

[lymereJ]

The multipliers are from the PRM-RM. So we probably would want to specify that we changed the approach if that's what we want to do.

[dmaddoxwhite]

How about adding an argument to the method to set the maximum number of sizing factor adjustment runs? If we limit to three, without an override, a user might get a model that fails due to unmet hours with no way to fix it.

[mdahlhausen]

@dmaddoxwhite in my experience, equipment size is rarely if ever the source of unmet hours. If one iteration doesn't fix it, eight iterations won't either. The issue is more likely to do with schedule alignment (e.g. 5% occupancy when the HVAC system is turned off), design conditions (extremes in the annual weather file very different from design conditions), or some weird VAV sizing issue. For many of these, even an HVAC systems with 10x the capacity isn't going to solve them.

Users will generate models that have baselines that fail the UMLH test. I'd rather they fail faster (3 iterations) than wait for 8 and wonder why it still didn't work. Alternatively, we can add some QAQC checks to point out basic issues (like 5% occupancy in zones when the HVAC system is off). I've already got methods in standards to check for those sorts of things.

[mdahlhausen]

looks fine; made some minor cleanup fixes and changed number of adjustments to 3. It may be worth adding some checking in here to pick out the cause of unmet hours, rather than just increasing sizing factors. The PRM RM includes some direction beyond sizing, but IMO it's not very specific or actionable.

[mdahlhausen]

@weilixu @lymereJ @dmaddoxwhite good to merge. Up to you if you want to edit the # of adjustments further or make that a user argument. It's also fine as is.

[weilixu]

@weilixu @lymereJ @dmaddoxwhite good to merge. Up to you if you want to edit the # of adjustments further or make that a user argument. It's also fine as is.

Thanks @mdahlhausen, we will update the number of iteration.
We also need to disable an optimal control in the PRM method. Will make a push shortly.