What you're thinking of doing is generally called a "parallel path" solution, which works okay if there's not much 2-D heat flow nor large thermal mass effects. This is typically how wood-frame walls in the US are modeled. i.e., with separate constructions for a stud and a non-stud portion, and a framing fraction indicating the percentage of wall for each. Unfortunately, this does not work so well for mass walls nor for metal-frame walls where there are both of these effects. If you look in the ASHRAE Handbook of Fundamentals 2009, Chapter 27, you'll find several methods to calculate equivalent one-dimensional U-values for masonry and metal frame walls, developed largely by folks at ORNL.
Of course, these are still steady-state approximations. To get the dynamic effects of 2-D heat flow in walls, one needs to calculate response factors or transfer functions, for which there have been numerous papers and Ph.D. theses, Ceylan and Myers 1980 and John Seem 1989 come to mind, but sad to say I've yet to see these methods integrated into general use building energy simulation programs (anyone correct me if I'm wrong). What I do know is that way back in the late 1980's, Zulfi Cumali led an effort to incorporate the Ceylan-Myers technique into DOE-2.1B for calculating foundation heat flows, but that effort never made it into the standard release version. In the early 1990's, I worked with a colleague then at LBNL to take the Ceylan-Myers technique out of Zulfi's DOE-2.1B and make it a standalone program WALFERFN (stands for Wall Finite Element Response Factor - New) to calculate equivalent one-dimensional RFs taking into account 2-D heat flow. WALFERFN was then used quite extensively to calculate a library of RFs for metal-frame and concrete walls for the California Energy Commission (1995). The report can be downloaded here.
Before anyone reminds me that EnergyPlus uses CTFs and not RFs, let me just say that the conversion from one to the other is extremely simple (see here ). However, there's still the question whether EnergyPlus permits reading CTFs from a file, or that it insists on doing everything itself, in which case WALFERFN would have to be inserted within EP, which would take months.
If you detect a sense of regret in this post, you're right. If there's anyone out there who would be interested to expand on or update WALFERFN, please let me know. In my mind, it's a well-proven piece of software with a solver that has yet to run into any problems. However, it's programmed in Fortran 77 and has only a text IO.
