Inconsistent hourly vs sub-hourly stochastic hot water schedules

[shorowit]

If I compare annual full load hours for stochastic schedules using 10-minute and 60-minute timesteps (e.g., here vs here), I get:

	Full load hours (60-min timestep)	Full load hours (10-min timestep)	Difference
occupants	6689	6707	0%
lighting_interior	2086	2077	0%
lighting_garage	2086	2077	0%
cooking_range	534	534	0%
dishwasher	213	213	0%
clothes_washer	134	105	22%
clothes_dryer	151	151	0%
ceiling_fan	3016	3009	0%
plug_loads_other	5388	5393	0%
plug_loads_tv	1517	1505	1%
hot_water_dishwasher	273	146	47%
hot_water_clothes_washer	346	154	56%
hot_water_fixtures	887	397	55%
hot_water_showers	380	142	63%

As you can see, the hot water and clothes washer values are very inconsistent. (Interestingly, the dishwasher is fine?)

This is impacting the results in #1502.

@rajeee @jmaguire1 @joseph-robertson

[rajeee]

I think the issue is that the schedules are always normalized so their full load hours cannot be calculated this way. The schedule values only represent when the equipment is running and their relative magnitude in relation to the peak. The value of 0.5 in the schedule_10_min.csv and schedules_1hour.csv doesn't represent the same magnitude due to the normalization (and the fact that 0.5 for the hourly schedule means the equipment consumes energy at 0.5 level for the whole hour, vs only for 10 minutes for the 10-min schedule.)
One easy comparison is to see how many non-zero schedule hours there are, and both files have 278 non-zero hours for clothes washer.

[jmaguire1]

Good point Rajendra. This might actually be more prominent for HW use in particular, because you'll end up with events that are only a few minutes long, whereas I think most appliances you'd expect to run for at least 15-20 minutes? So the difference in the normalization factor for DHW could be a lot larger than it is for other end uses. At least I suspect that's the case.

[shorowit]

Ah, I forgot that the power design level (or peak flow rate) is not constant, but rather calculated from the desired annual energy use and number of hours in the schedule. So these schedules do seem correct, in that you'd expect higher peaks (and therefore fewer schedule hours) when using sub-hourly timesteps. For other appliances (dishwashers, cooking ranges, etc.), they have long event durations at the peak, so the peak does not change between hourly and sub-hourly timesteps.

Okay, I'm satisfied here that there isn't an issue. Closing it. Thanks for the investigation and discussion on Teams, @rajeee.

Now that I understand, I'll need to revisit the unmet water heating load PR and think about whether the implementation there is correct...