We present a multiobjective optimization framework to evaluate the effects of comfort relaxation on
the energy demands of buildings. This work is motivated by recent interest in understanding demand
elasticity available for real-time electricity market operations and demand response events. We analyze
the flexibility provided by an economics-based control architecture that directly minimizes total
energy and by a traditional tracking control system that minimizes deviations from reference temperature
and relative humidity set-points. Our study provides the following insights: (i) using percentage
mean vote (PMV) and predicted percentage dissatisfied (PPD) constraints within an economics-based
system consistently gives the most flexibility as comfort is relaxed, (ii) using PMV and PPD penalization
objectives results in high comfort volatility,(iii) using temperature and relative humidity bounds severely
overestimates flexibility, and (iv) tracking control offers limited flexibility even if used with optimal
set-back conditions. We present a strategy to approximate nonlinear comfort regions using linear polyhedral
regions, and we demonstrate that this reduces the computational complexity of optimal control
formulation