I get the answer from eQUEST-USER list as following:

For VRF systems, I’ve seen direction to use PSZ, PVVT, and PTAC. PVVT is the approach endorsed/outlined by the Oregon energy trust in a powerpoint PDF that has made the rounds here on the lists a few times, and as such is probably “most suggested.” PVVT by distinction can model variable/multispeed fan control at the terminal level, which may or may not reflect your actual equipment (depends on make/model). Fan control can also be made “constant” in that regard if need be, as such this may be the most flexible solution. PSZ stands out from this trio as a zonal system type – this fact may make it an easier system to consider in relation to a model or baseline you’ve already substantially developed with a zonal system type.

There is more than one right answer, and it’s worth noting/acknowledging VRF is fundamentally different in operation than any system type currently within eQuest.

Review of LG and Mitsubishi’s suggested practices / guidance shows some slight contradictions. Both notably prescribe system type PSZ, to start, but Mitsubishi explicitly mentions PVVT and PTAC as viable alternatives as well. Somewhere along the way I recall reading that energy recovery elements may be more easily accounted for with PSZ over PVVT… can’t recall the specific source. LG by has produced a library of (air-cooled) curves specifically for use with modeling variable speed compressors. Mitsubishi in contrast (perhaps, in response… also for air-cooled VRF) asserts it isn’t critical to explicitly model variable speed compressors to approximate VRF within eQuest, rather that the PLR/EIR/Capacity curves furnished by the manufacturer can, together, adequately address performance.

As I understand it, any chosen approach to VRF in eQuest involves “post-processing” hourly reports to quantify and sum “free” heating/cooling energies. There is some variance in guidance (between the energy trust of OR and Mitsubishi) on how to sum those free heating/cooling energies... both methods appear conservative to my understanding of how VRF works (summing each hour’s energies in isolation of the preceding hourly results), but then I am certainly no expert.

~Nick