Control-Oriented, Data-Driven Models of Thermal Dynamics

We investigate data-driven, simple-to-implement residential environmental models that can serve as the basis for energy saving algorithms in both retrofits and new designs of buildings. Despite nonlinearity underlying dynamics, using Koopman operator theory framework this study we show a linear second order model embedding, captures physics occur inside single or multi zone space does well when compared with data simulated EnergyPlus. This class has low complexity. their parameters have physical significance large-scale dynamics building are correlated to concepts such thermal mass. consequences changing mass on behavior system provide best practice design suggestions.