Unexplained start date sensitivity in E+ Simulation
For the attached model, differences in monthly total electric energy consumption were observed when the start date of the simulation was changed. For the plot below, the end date (and everything else) were held constant.
More pronounced differences between runs (w/ different start dates and the same end date) were observed when AMY weather was utilized.
Has anyone experienced this sensitivity before? I also increased the minimum # of warmup days to 200; no decrease in sensitivity to start date was observed. From the *.eio file, it was confirmed that the autosize results were the same regardless of start date.
UPDATE: Also performed an AMY weather comparison between starting the SIM in July(full year) and September(partial year), with both SIMs ending in June. As shown below, the difference in total electric energy usage persisted throughout the SIM duration. Majority of the difference observed was in heating and cooling electricity.
When I started reading, I was going to suggest warmup days, but you have certainly looked at that aspect. Are you doing sizing with these runs/are your sizing parameters and design conditions the same for all scenarios?
I'd plot the end uses at detailed frequency and see which one(s) show the divergence. My guess is that it will be heating or cooling energy, and that this is the result of the difference in starting conditions of something with inertia, like a plant loop's water temperature or the mass of the building itself. If that's the case, you could plot the conditions of that thing and see how the mid-run conditions match up to the result of the warmup period. Also, you probably know this, but 200 warmup days means run the first day over and over 200 times, not go back in time 200 days and run them.
I don't think it's odd. Your weekday vs weekend days will be different if you change the start date of the simulation. So you are running the building during different weather conditions. Have you got the first date of the simulation fixed to Sunday?
@Liam Using the full set (7 total) of sizing days and kept them consistent between all simulations; I also checked the EIO file and confirmed that the autosizing values were consistent between runs
@TomB Thanks for the feedback; for both TMY and AMY weather runs I adjusted the start day accordingly; it seems like odd behavior for an actual building, thus the search for an explanation with the simulation; this simulation is based on an existing building and we've already verified the general simulation behavior vs. sub-meters and utility data