This paper presents a physical-statistical radio channel power delay profiles model for room-to-room communication systems combining the room electromagnetic theory for modeling deterministic channel components with a geometry-based stochastic channel model with time-variant statistics for modeling stochastic components. The deterministic channel component, i.e., mean power delay spectrum, is comprised of specularly reflected paths plus diffuse components due to scattering and diffraction. The specular components are modeled with a set Dirac function, whereas the diffuse components modeling approach is a room electromagnetic theory-based model. Dynamic indoor communication channels are characterized by a non-stationary time- and delay-fading process due to changes in the environment. We analyze and model the time-delay variability of channels using $K$ -factor for small-scale variations and the $t$ -location scale distribution parameters for large-scale variations. It turns out that these parameters cannot be assumed to be constant in time and delay. After modeling of time-delay variations of the first order statistics, we generate channel realizations with appropriate second order statistics. As the result, the presented model enables to describe the evolution of the power delay profile in the time domain.