Hello modeling community,

I am at long last announcing the publication of my dissertation and associated USGBC-funded research that investigates the application of model predictive control to mixed mode buildings, examines the optimality of typical control schemes for these facilities, and explores the use of data mining techniques to automate the extraction of near-optimal control rules from offline model predictive control results (in other words, training simplified decision models to operate a building in a near-optimal manner based on heavy offline computation with detailed EnergyPlus models). Research was conducted with oversight from my advisor, Gregor Henze at the University of Colorado Boulder's Building Systems Program. You may find the work particularly interesting if you have explored the related research of Brian Coffey (also recently posted to this forum).

The results are available through a few publications listed below:
Dissertation, Offline Model Predictive Control of Mixed Mode Buildings for Near-Optimal Supervisory Control Strategy Development. A thorough treatment of both the MPC and rule extraction processes, with experimental validation and associated source code.
USGBC final report, HVAC Control Algorithms for Mixed Mode Buildings. A slightly more accessible report that ignores the rule extraction angle and instead benchmarks existing mixed mode control strategies against MPC. Simplified, near-optimal heuristics are analyzed.
For those with access, you might be interested in the following journal publications: Extraction of supervisory building control rules from model predictive control of windows in a mixed mode building (JBPS, 2012); Model predictive control of mixed mode buildings with rule extraction (Building and Environment, 2011). One more is in the works on experimental validation.

I am more than happy to discuss the work and share associated rule extraction code with interested parties. Hoping this can be of use to some of you.

Best regards,

Peter May-Ostendorp, PhD, LEED AP