Sam's profile - activity

High-volume != accurate. I assume that these simulations are for jumping through LEED hoops and not design? As in a below comment, actual vs predicted EUI:

www2.buildinggreen.com/blogs/lies-damn-lies-and-another-look-leed-energy-efficiency

What are we actually simulating then with all this OS detail? Measured results don't provide such evidence. I worry for those who think they can simulate their way to good design...scripting isn't the problem, it's the lack of understanding realities.

Absolute is all that matters: one building is built and does it perform as predicted? Comparative in the sense of design alternatives? Systems-wise, any design engineer worth their salt knows what works where and why. Here is the scatterplot of actual vs. predicted EUI (CBI not Berkeley) on buildings where supposedly there was an active attempt at simulation accuracy. http://www2.buildinggreen.com/blogs/l...

TRNSYS if you want to simulate "how an engineer thinks," and E+ if you want a 1990's "object-oriented" approach where you can't model/simulate that object in isolation. Modelica for the 21st century.

If not, I want to slowly build a code base that allows for easy, straightforward sensitivity studies for client engagement.

I think "accuracy" is the key word. I don't think building energy modelers are incentivized to produce accurate models other than on aggregate levels. Plus or minus 15% energy consumption (whole building) is "good enough" and whose going to check anyway. Have you seen any studies comparing actual vs. simulated energy consumption? I think Berkeley showed the shot-gun blast that is measured vs. predicted EUI. Some will say, well the occupant usage cannot be predicted yadda yadda yadda. Well, than beyond basic design decisions, what are we simulating? Bells and whistles, kids. OS, pile them on!

In my opinion this is the major problem with EnergyPlus. You can't really simulate anything in isolation, so instead of "debugging" subsystem by subsystem you have to simulate the whole building and hope there are no major errors. Poor design choice, but especially a problem when you're trying to calibrate subsystems to real data.