[Help-mcsim] MCSim and hysteresis?

[Bill Harris]

Has anyone here used MCSim to model a hysteresis process?  Think of the
energy used by a building as a control system (thermostat) turns on and
off.  Instead of a 0.25-0.5 degree C hysteresis band, as you might see
in a home system, think of a significant band -- perhaps a 5-10 degree
hysteresis band.  That's not quite the problem, but I hope it makes it
understandable. 

What I'd want is something like two linear (or perhaps nonlinear)
equations describing energy use per unit time as a function of outdoor
air temperature, one for each state of the system (there could be more
than two states, of course).  The model would need to estimate the
temperatures at which the system switched between the states, and I'd
want to be able to feed it a new path of temperatures and have it
calculate the distribution of predicted energy use per unit time for
that path.

I've looked at hidden Markov models, but I'm not sure the assumptions of
HMMs quite fit, and I've not made it work (R's depmixS4, for example).
It does nicely separate the observations and transitions, though.

I've looked at R's cusp, but I don't yet understand how to model the
observations and their uncertainty.

I've considered relearning Kalman filtering but haven't yet.

Any ideas on how to set up the model in MCSim?  Any examples?

Thanks,

Bill
-- 
Bill Harris                  http://makingsense.facilitatedsystems.com/
Facilitated Systems                              Everett, WA 98208 USA
http://www.facilitatedsystems.com/               phone: +1 425 374-1845