Energy performance assessment of occupied buildings using model identification techniques

Abstract Model identification allows the assessment of energy performance of buildings based on measured data. However, model identification of occupied buildings is difficult because the disturbances introduced by the occupants are usually not measured. For office buildings, we propose a method to estimate the occupancy in function of electrical energy consumption. To determine the complexity of the model, SVD is used for the number of thermal zones and statistical analysis is used for the model order. The estimated model was evaluated by simulation, consistency of input–output behaviour and uncertainty analysis. The RMS of the simulation error is less than 0.7 °C for the training data set. The proposed method allows calculating macroscopic parameters of the building (such as thermal loss and time constants), showing the relative weights for heat losses and estimation of energy savings by heating control.