This paper presents a new methodology for modeling building energy performance that addresses some important limitations of building simulation. This new methodology develops a physical model for accurately predicting indoor environmental conditions and energy consumption by selecting best match parameters and variables. The innovative aspect of the proposed methodology is the introduction of open and closed loop system approaches to dynamically model the complex interaction of factors that contribute to building thermal performance and their uncertainties. This allows simultaneous tracking of both the lead and lag times between heating excitations and indoor thermal responses to account for their mutually excitatory interaction. The model system is solved using a Laplace transform technique, with an explicit solution that includes physical and generalized parameters calibrated by measurements. Singular value decomposition techniques are applied to further determine the model variables for the best approximation using lower dimensions. As a result the model complexity and the model parameters and variables are minimized while still preserving the physical meaning of the model. A careful, detailed validation and assessment of the model performance is conducted using a case study of a dance hall at a swim center (R2 > 0.9). A further validation of the model is also undertaken by assessing its forecasting capability against benchmark persistence models. The proposed model outperforms the benchmarks especially over longer time horizons. The methodology is useful in developing a minimal but comprehensive and accurate energy performance physical model which can reliably capture the dynamics of building thermal and energy performance. The proposed method can serve the needs of prediction and control applications in a wide variety of building types and can be incorporated into the most commonly used simulation models.