Over 40% of the energy used in most cities is spent on heating or cooling buildings. Buildings do change dynamically in terms of their configuration, usage and performance. This calls for an online system which can monitor their performance in real time. A critical component of such a system is a well calibrated model of the facility that accurately predicts the actual facility's energy behavior. This paper discusses a novel methodology that allows modeling the energy consumption of a facility just with one energy consumption meter. The methodology does not require additional sensors of any kind. The proposed method leverages statistical analytics and evolutionary search techniques to calibrate the energy model of a facility. We present the methodology and a real world case study of a facility (in UK) in this paper.