This paper presents a methodology for parameter estimation of a suitable model for energy management services in office and apartment settings. The objective of this work is to identify model structures and tune parameters to fit recorded data. Once these parameters have been identified , the model will support energy services (prediction, explanation). Grey-box models are proposed to estimate the temperature and CO$_{2}$ concentration based only on few sensors. Then, two different estimation parameter methods have been applied relying on a descent algorithm or a genetic one. Different structures have been compared. Finally, the resulting model has been applied to two case studies: an office and an apartment.