Posted By toddm on 21 Aug 2011 01:02 PM
Sheesh. It's like the movie Ground Hog Day. Yes, ColoICF, a concrete home will cool slower than a low-mass home in same manner that a five gallon bucket of boiling water will stay hot longer than a one gallon bucket. (Analogy not perfect because the bigger bucket has more surface area radiating heat.)
but put them back on the stove and guess what? It takes the same number of btus to reheat each to boiling (again, assuming equivalent surface area.) You haven't saved nearly the energy that you think you have. Here is a link to an ornl report on the joint icfa/ornl test houses in loudon tn completed in 2001. http://www.eere.energy.gov/buildings/building_america/pdfs/db/30962.pdf
The cliff notes: the thermal mass effect of ICF in tennessee is in the 8-10 percent range. The ICF house is 5-10 percent tighter. The jury is still out on ColoICF's earth coupling theory. I assume it is still out judging by how deeply the ICFA has buried this study, relying instead on BS research.
But that doesn' mean that we can't learn something from the ICFA. Hey, my analysis shows that I am 40-50 percent handsomer than the gents at the Golden Acres Nursing Home. I am still waiting for the babes for each arm.
The more detailed (and telling) synopsis of that study lives
here.
Total energy savings of ICF vs. the code-min 2x4 house was 7.5% (unoccupied) to 9.2% (occupied.)
Scrolling on down to Table 1, p3 (of pdf pagination), the BIGGEST difference in the features of the two houses was the whole-wall R values of the exterior walls, where the ICF was R15, compared to R10.6 for the code-min house.
What- a house built with R15 walls uses less energy than one with R10.6 walls!?! Who knew? 
The code-min house had ~10% higher air infiltration rates as well. (Table 2, p.5) and slightly higher duct leakage (Table 1), which will account for some of the difference, maybe 2%, but that's not nearly as much as ~30% lower whole-wall R values. A 2x6 wall with R21 batts also delivers about R15 whole-wall, which would have made for a non-straw-man type of comparison (but that's above code-min in some places.)
One experiment proves nothing, of course, but energy use for both houses tracked the DOE2 modeling fairly well, validating the model, by which they predicted percentage performance deltas between the two house for a number of different locations & climates (Table 4, p.8), none of which hit double-digits.
The costs of air-sealing and adding an inch of foam to the code
min house or go to 2x6 construction to bring it's wall U-values in line with that of the ICF house is a lot less than the cost of building with concrete vs. code-min stick built, and the difference in energy use difference at equal-R would be well under the 7.5%-9.2% enhancement measured in this experiment.
Of course there are many good, even GREAT reasons to build out of concrete- it's a fantastically durable material, but (as I've said before), energy use can't be the driving rationale for going with ICF. Yes, you'll get some boost from the mass cutting into peak loads, but in most cases that won't affect the sizing of the mechanical systems, but it's still the case that R-values & air tightness, not thermal mass, will continue to dominate the energy use numbers. (The exception is when thermal mass is designed in as part of the passive-solar aspects of a house, and not isolated from the conditioned space by EPS as it is with ICF.) There's no magical thermal differences between EPS on concrete compared to EPS/other-R on stick-built assemblies, and the higher the total R, the less the concrete mass really matters.
In this
Boulder CO model showing a large difference in peak and average loads they're comparing an R12.8 stick-built cube to an R11.1 mass-wall cube, neither of which would in fact meet code in Boulder, nor are they indicative of how anyone actually builds. (Can you even GET an R11 ICF?) They don't specify how one could even build as low-mass as less than 1BTU/hr/ft
2 per degree F, as spelled out in the model either. The ORNL testing on actual assebled houses is much more realistic, if on the- low-end for R values.